Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
From ISRO Scientist to $25M Deep Tech Founder | Prateep Basu (SatSure)
0 Plays2 seconds ago
This episode with Prateep Basu, Co-founder and CEO of SatSure, is the deep-tech founder story India's startup ecosystem has been waiting for.
Prateep Basu left a career building propulsion systems for India's GSLV MK-III rocket at ISRO to ask a deceptively simple question - why do urban Indians get 10 loan offers a day on WhatsApp while farmers wait a month for a single approval? The answer became SatSure, a Bengaluru-based Earth intelligence company that uses satellite imagery, AI, and government land records to deliver alternate credit scores for farmers, monitor crop health across bank portfolios, and help airports, insurers, and FMCG companies make smarter decisions from space. Bootstrapped for four and a half years before raising $25 million across multiple rounds, SatSure now monitors 1.95 lakh villages and has analysed over 2.1 million farmer plots.
In this candid, wide-ranging conversation with host Akshay Datt, Prateep breaks down the physics of 40-pixel crop detection, explains why algorithms are never the moat, and reveals the strategic logic behind SatSure's audacious zero-bid for India's first private national satellite constellation. He also shares why AI is an accelerant, not a threat, for deep-tech companies that have built genuine domain depth.
What you will learn in this episode:
👉How SatSure built an alternate credit score for farmers using satellite crop imagery, land boundary data, and historical yield analysis, collapsing a 30-day loan process to under 30 minutes and creating a product that banks like ICICI and IDFC are paying for at scale
👉Why Prateep believes the algorithm is never the moat, and how SatSure's real competitive advantage is the nuanced translation of banking business processes into product and model design, something no open-source GitHub repository can replicate
👉The full story of SatSure's 4.5-year bootstrap, from a letter written by a young MP in Srikakulam to winning a Gates Foundation-backed challenge and securing a five-crore purchase order from the Andhra Pradesh government, all before raising a single rupee of institutional capital
👉How India's Digital Public Infrastructure, including AgriStack, the Unified Lending Interface, and digital land records, is assembling the exact pipeline that makes SatSure's products commercially unstoppable at national scale
👉The strategic logic behind Allied Orbits, the consortium of Pixxel, SatSure, Dhruva Space, and PierSight that won India's first private satellite constellation contract, and why the consortium chose to refuse government funding in exchange for full global commercialisation rights
#DeepTechStartup #StartupIndia #EarthIntelligence #RemoteSensing #SatelliteImagery
Disclaimer: The views expressed are those of the speaker, not necessarily the channel
Recommended
Transcript
Satellite vs Drone Intelligence
00:00:00
Speaker
Intelligence from satellite doesn't compete with any other sensor. The drone versus satellite argument itself is stupid. they We compete with human intelligence. China would have about 300 satellites which are doing the earth imaging. India has what, eight?
00:00:16
Speaker
The government of India are already spending close to 800 crores buying earth observation data and intelligence.
India's Investment in Earth Observation
00:00:22
Speaker
Government today prioritizes Ability to act fast, seeing the deforestation triggers, deforestation program, what is the carbon stock and how
Founding of Satsure by Prateep Basu
00:00:30
Speaker
it's moving. Prateep Basu is an ex-ISRO scientist and the founder of Satsure. Satsure is an earth observation company. It helps banks, governments, insurers to make decisions using data from space about things like lending, farming, forestry, infrastructure and
Adapting Algorithms for India
00:00:47
Speaker
so on. Algorithms come dime a dozen. Making it work in the Indian conditions to the specifications of what your user wants. That requires a very ah nuanced translation of the business process embedded into your model design or product
Satsure's Position and Offerings
00:01:04
Speaker
design. What are the big untapped opportunities which you haven't even touched so far?
00:01:16
Speaker
Prateep, welcome to the Founder Thesis podcast. ah My first question to you, what's the right industry ah for Satshore? Is it space tech? Because you're not the hardcore space tech company like, say, a pixel. Is it like...
00:01:33
Speaker
um like Is it like in the in the data space? Is it SaaS? What is it? like How do you categorize it? I would want to know what is the definition of hardcore space tech at another point. I'm saying like people who launch rockets are typically seen as more space tech-y-ish. But I could be wrong, man. I'm an outsider. So don't hold it against me. you You
Intelligence-Based Business Solutions
00:01:59
Speaker
tell me. What's the right way to classify Satchelor?
00:02:02
Speaker
I think my company is at that nice event horizon, which brings outsiders and makes them insiders, you know. So we are at the intersection of what we call as space yeah space technology, earth observation and artificial intelligence. So we we nicely coined ourselves as an earth intelligence company.
00:02:23
Speaker
oh Okay. So what are you selling to your customers? You're selling intelligence. Well, we are selling outcomes, to be honest, ah outcomes that are derived using ah imagery that is ah you know translated into initially the intelligence intelligence from the image imagery and integrated with non imagery data sets that you know we continuously integrate. It depends on use case to use case. And then that's how you deliver a and know business outcome, which has a certain value associated with it. ah so but What does it mean when you say you're selling outcome?
00:02:59
Speaker
I like outcome would mean that you I mean, my understanding of outcome is as follows. So, for example, if somebody is running a marketing agency and that for them to sell outcome would mean that, OK, I'm giving you leads. OK, I'm doing lead generation for you, for example. So how how are you selling outcome? Just break it down in a more layman friendly way.
Lead Generation for Agribusinesses
00:03:21
Speaker
That's a great example. I'm going to say lead generation. You're going to take this example of lead generation only. So as an as as take a use case with agribusinesses and banks, ah we do enable them to do lead generation. Now how? For them, the lead generation is the factor of following things. Number one, they need to ah They need to know whether the leads are close to their local branches or retailers, which is physical presence.
00:03:48
Speaker
And it should be within the catchment area, like the area that they can cover physically. Second, they need to filter but the the leads by whether they these farmers are growing the kind of crops for which they are selling the product like a pesticide
Interactive Data Platform
00:04:04
Speaker
or a seed. or you know ah having active land availability ownership which in the case of finance so merging all of this information uh allows us to provide that outcome that this is your best list of the top 20 farmers in each village as an example uh so is it uh
00:04:29
Speaker
it does is there someone in your team who's constantly giving them answers to questions that they're asking or how how are you doing this like say top 20 leads in your in near a branch in some village you know so is this a question which comes in or is it productized that every branch has access to some sort of a ah interface where they can log in and search for customers and apply filters and say okay only within five kilometer radius with landholding of more than one acre, show me all farmers and then you
00:05:00
Speaker
pop that up as a result absolutely so this is completely platformized uh where our data lakes are integrated there are certain filters uh and uh it allows uh our customers to upload their uh data sheets in the form of you know csvs whatever uh and uh they can slice and dice and visualize the information and down download that from the platform indeed it did even create an api to integrated into their salesforce, SAP or whatever you know process tool
Data Accuracy and Validation
00:05:30
Speaker
they are using. So ah that's the extent to which we go to make the consumption experience easy, to make the data that they are making these decisions on trustworthy, where it's not just like, okay, Sasha has published some data with some crop. How are we providing the validation? that How accurate is the information? Not everything can be ground tested. like you know Each and every village, I cannot have this information collected. right So, ah auxiliary data support, which means that you know you have a validation tab and ah you see ah what's the patterns, etc. etc.
00:06:09
Speaker
Historically, what is the aggregated numbers government has published ah in the past, some of the ground truth-led information in and around that area. So this this is a complete change in the way ah imagery from space goes into intelligent data-driven decision-making for
Streamlining Loan Processes for Farmers
00:06:28
Speaker
enterprises. That's what at Satshwar we do.
00:06:31
Speaker
OK, give me like a real world example with, let's say, a financial institution. but what What is the problem statement they came to you with? And what does the product look like for them? How are they using it? What decision making does it lead to?
00:06:48
Speaker
Right. So the example I gave is also a real world example only. ah Maybe mean I'll give you and another another one. um So one of the first problems that we ah attempted to solve was ah shortening the process of ah sanctioning a loan to a farmer um and our customers are banks. now For the banks, the process is broken because ah there is a need to physically visit ah the farmers' residence, the farm, get a third party allegiance, identify if the land belongs to the farmer, it's fragmented, how much portion of it belongs to the farmer, whether he has the experience of growing the crop for which he is asking for a loan, ah what's the availability of water, ah you know, how what is the region like it up for drought floods availability to market there are too many factors and not everyone is uh uh you know the expert in agriculture markets uh within the within the banks so uh we saw this as a as a very interesting
00:07:54
Speaker
an impactful problem to solve Akshay because ah ah it seemed a bit unfair that people like you and me ah would get 10 times a day WhatsApp messages and calls from all kinds of banks, NVSCs on what loan eligibility we have based on our civil score or CRIF score. Why can't that be there for farmers?
00:08:15
Speaker
That's the gap we were filling. That in order to get the farmers have the same user experience as an urban user like us, This is the need of data, the land information, the ownership information, the crop information, and the climate risk
Satellite Imagery's Role in Risk Assessment
00:08:30
Speaker
information. So by but stitching it all together in the form of an alternate credit score report, just like you download a CYBEL or a CRIF report, ah we essentially shortened the process of 30 days to a few minutes, like 10 to 30 minutes now.
00:08:44
Speaker
Wow. um And this you're saying is not just based on ah EO or Earth Observation data, which comes from satellites, but you're also using multiple other sources of data.
00:08:56
Speaker
Absolutely. The interesting thing about every ah use case that we ah build is that Despite all the other data that needs to bring together needs to be brought together for making that digital journey, ah without the satellite image intelligence piece, it would never work.
00:09:17
Speaker
So in this case, the ability to correctly identify the cropping patterns, the production patterns, the the irrigation patterns, at a plot level can never be ah done by any other source ah in order to meet that few minutes timeline. You cannot send a person or a drone to do that when you receive a request.
00:09:40
Speaker
And especially when you are measuring risk. Risk is all about history. You cannot send again all other data collection tools back in time to collect that data. ah How did you build this product for this bank?
00:09:52
Speaker
Because just purely by virtue of having access to ah Earth observation data from satellites is not going to solve the bank problem. right ah Just take me through a journey of actually building out the solution for the bank.
Banking Pilots and Challenges
00:10:05
Speaker
So look, our backgrounds, when I say our, um myself, my co-founder and founding team ah was a lot from the space industry side. We used to be working in ISRO, so different, different domains. We know at least ski what can be done using satellite data. ah And the problem of financial inclusion ah was was ah big enough for us to take a stab at it, which means we went and sold the problem first to the banks. We didn't sell a solution. We didn't have a solution. So ah as the banks, some of the early banks, private sector the banks like ICIC, HDFC bought into the thing, they said, hey, this looks exciting, actually, if you can solve it. So we don't know whether you can solve it because we can't evaluate your technical capability. But what you're saying, ah if you do it, this is transformational. So we ah entered into pilot agreements 2019 to 2021 with a couple of banks and used that opportunity to learn about the business process, different user personas, go to the field, see how a life with the credit ah managers you know day looks like, and essentially created the full product
00:11:13
Speaker
out of the initial capability that we sold in the form of the problem definition. OK, interesting. I want like slightly more zoomed in view of how you stitched all of these pieces together. So you understood what the problem was, what the need for data was. Then how did you stitch it all together?
00:11:34
Speaker
Right. So there were two very important pieces of it. ah One is how can we ah improve the confidence on the ah boundaries of the firm and whom it belongs to. okay These are two different data sets. We have to integrate these two. The second one ah is how can we build the trust on the ah
00:12:05
Speaker
the precision with which we ah identify the cropping patterns because the loan product is linked to the area of the land and the type of the crop. These two are the most important part. It defines the scale of finance, like how much amount of money you ah provide.
00:12:20
Speaker
So like someone who's growing... wheat will be eligible for a certain loan value because you know that wheat will be sold at a certain like there is a certain uh like the the wheat is worth something there is a market price for wheat therefore the loan can be given for a certain amount okay understood correct so these two are the most important part everything else were good information to have which supports the decision on based on these two.
00:12:48
Speaker
Now the boundary ownership piece is something that that as the government was executing its digital land record ah program ah that started becoming more and more public data sets. And governments also ah figured out how they could publish this information through the ah digital public infrastructure like AgriStack.
Incremental Development and Partnerships
00:13:09
Speaker
While the crop identification issue is fairly more complicated in India because they are very small farms and there is too much heterogeneity. Heterogeneity meaning that adjacent farms will have different crops and you don't have large swaths of same crop like you see in the US etc.
00:13:27
Speaker
um which meant that we needed a lot more field-collected data for training the models which are used to identify ah the crops unlike in us and that's not something which was publicly available unlike again the us ah so i would say step by step we went about it meaning that ah bank would say that hey i cannot use uh your service at full country scale because uh you know
00:14:01
Speaker
I need to draw confidence on all the crops that we fund. and We cannot do all the crops we fund. So we said, okay, let's filter um top five, six crops that constitutes 60% or 70% of your portfolio. So first we developed ah the capabilities on that where a precision of 90 to 95% is reached at a farm level across, let's say, the major lending states ah in in the country. ah which What are those top five, six crops?
00:14:32
Speaker
So there will be ah you know paddy, wheat, cotton, soya bean, sugar cane, oil seeds like mustard and stuff like that, maize. So um we we developed those capabilities by indulging into both the ah what we call as ground truth, meaning going to the field, drawing a polygon, taking a picture, marking that, OK, this is the crop we saw and all that, while also ah you know doing a bit of data barter system with early customers in the piloting phase that they use their field force and providers data and we give them a discount on the final intelligence on the imagery. So like that, we developed a a repository of this data that kept strengthening the algorithm algorithms on each of these crops to a point where the banks
00:15:24
Speaker
ah could say that okay ah now we have seen what are the changes in the process because you have enabled a digital journey for land I'm happy with it am I able to take a credit decision yet on your crop no but uh not now and until now but now this so let's say uh trust is built on your precision now we can take decisions this goes into our risk and policy uh then last part is how you know we built a score scoring mechanism now how the score holds against the real defaults uh that the banks has faced back test that together so that to like eventually a
00:16:01
Speaker
the the decision to lend or not lend is a digital uh process where uh let's say our score start score uh of 780 for a you know one hectare of wheat uh one lakh rupee auto approval is set meaning no manual intervention to do checks required so this kind of ninety out of 800 and Out of 1,000.
00:16:27
Speaker
Out of 1,000. OK. Yeah. So this kind of product definition started getting built as the layers after layers of information got added, validated, and then gone through the process.
Credit Scoring System for Farmers
00:16:38
Speaker
So it's a long, arduous journey to you know bring change.
00:16:44
Speaker
um OK. So the purpose of a score is essentially for a bank to have confidence that I will get my money back. Right. Yeah. What is the score based on? So you're saying that the what you understood is that the bank needs to make sure that the guy who's applying for the loan is actually the owner, A, and B, the boundary which he's claiming that this particular area is my land, is actually his land, and C, that he has experience of growing that crop.
00:17:11
Speaker
Right. So the score is essentially providing the volatility of the ah income of the farmer, right? ah that predicts the potential of default ah moving forward. So, it's a predictive score based on historical patterns analysis.
00:17:29
Speaker
but What patterns? Like what all crop he's grown? Is that how you're analyzing or...? Because if you know the crop and the production yield here and you have ingested also the market pricing data because either it is governed by our states and ministry like ah ah the minimum support pricing or the you know cash crops are traded the market. So you have the pricing data. So now you you can make a upper limit, lower limit of what is the revenue that has been made. And there is a certain cost of cultivat ah cultivation associated with each, let's say, crop per hectare. And you know what's the you like you know what's the availability of cash and the movement of that season on season for the farmer.
Advancements in Satellite Imagery Technology
00:18:12
Speaker
Okay. Okay. Okay. So based on how much he has been growing, what has been his output? And his output you can see because you have historical satellite images.
00:18:23
Speaker
So historical satellite images are generally available with high fidelity. ah for like For how many years back like do you get? like See, a company like us wouldn't have existed only ah because of this problem. And it because and ah not until
00:18:43
Speaker
satellite imagery of 10 meter resolution, which means one pixel is 10 meter by 10 meters. So in one acre, you have 40 pixels at least. ah This was not this kind of imagery was not openly available. So European Space Agency made that ah available.
Open Data Utilization
00:18:59
Speaker
And today, quite a lot of space agencies now publish data with high fidelity across the globe as a part of, you know,
00:19:09
Speaker
their service in the form of digital public good. So a company like us uses that and adds value on top of it. So we, for example, can analyze any point on the Earth using this particular set of imagery on a five-day basis.
00:19:26
Speaker
So one acre is 40 pixels. And so a 40-pixel image, your algorithm is able to predict whether it's wheat or rice.
00:19:37
Speaker
That's the secret sauce, yes. Okay, that's pretty amazing. With just 40 pixels, that's phenomenal. that's Okay, amazing. ah Got it. Okay. And how much data did you need to train your algorithm? Because you said you need validation through field data, like like the algorithm would say this is wheat, but you also need to send a human being and check is it actually wheat and then tell the algorithm if it's right or wrong to train it. Okay.
00:20:05
Speaker
how hard was that process or maybe you're it wasn't a lot of data that was required if we look at it now compared to so much of the AGI work that's going on uh maybe we would have used about you know 300 uh thousand data points dispersed across because you have to smart
Algorithm Tailoring for Indian Conditions
00:20:29
Speaker
about it. It's such a huge country. ah And in terms of agriculture, it is bigger than but you know Australia, US also. ah
00:20:36
Speaker
It's very arable. So ah statistically sampling ah the data, like the points where you want to collect data, I think that was the reason where we could reduce the number of this field collection over a period of time.
00:20:52
Speaker
okay and your bank partners helped you like their field force would also provide you human verification of sometimes and those are not also you know so so easy because the quality of the data collection would be poor a lot right okay okay okay so ah okay you so you took ah the satellite imagery, which is publicly available and free of cost. And did you also have to pay for any of the data? Or most of this is built on like open source data, if I can use that term.
00:21:26
Speaker
Most of it is based on open source data, weather and satellite images, et cetera. OK. So essentially, the moat for you is the algorithm.
00:21:39
Speaker
the ability to take open source data, put various layers of open source data, one on top of another, map those layers with the customer requirement because a bank has a very specific requirement of what crop is being grown and so on, and then map it with the government land records, and then throw out a score.
00:21:58
Speaker
That is the note. I think the ah that's it's a bit oversimplification, I would say, because so algorithms come dime a dozen. ah And that also, you know, ah way too many go to GitHub and say, OK, I want a crop classification model on open data, you will get it. But okay making it work in the Indian conditions ah to the specifications of what ah let's say ah your user wants reliably consistently and adding the decision context layer because that final thing is an outcome, decision taken, right?
Integrating Business Processes
00:22:40
Speaker
ah that That requires ah
00:22:44
Speaker
very a nuanced translation of the business process uh embedded into your uh process like your your model design your product design uh and uh hence it becomes a lot more than just the algorithm ah open okay so business context uh specific uh like the ability to come up the credit score is not just An algorithm, but it's also the business context, essentially from what?
00:23:17
Speaker
Absolutely. Okay. okay Okay. Okay.
AgRi Stack and Satsure Integration
00:23:21
Speaker
Understood. Very interesting. You spoke of an Agri stack. What is this Agri stack? I was not aware India had an Agri stack.
00:23:28
Speaker
Oh, yeah, Agistack is a digital public infrastructure that central government started working, I think, around 2021. And yeah, it got operationalized late 2023 with Uttar Pradesh state being the first state to implement it ah for improving the governance and know delivery of all the farmer benefit schemes, etc.
00:23:50
Speaker
But what is the data that what is it exactly like? ah So what Agustad did was ah it brought together ah government data that's either been collected before or is as a process continuously collected. Things like area zone, production, uh information like boundaries revenue record etc soil information uh you know subsidy how much amount of fertilizer uh subsidy has been given city subsidy has been given all of this information got aggregated and cleaned up into one single uh repository of data which is then made ah available via apis uh to different uh
00:24:34
Speaker
kinds of organizations that provide services to farmers uh with a consent layer of accessing that data the consent being from the farmer themselves okay amazing that is amazing I was not aware that we have such an advanced that's pretty amazing uh how How does the consent from the farmer come? Like they get an SMS or something or they have to share an OTP.
00:24:59
Speaker
Correct. Not OTP, just SMS that this is the, ah like someone is wanting to access your personal information. See, there is information at multiple layers. Up to a administrative layer information that is okay, open, they can...
00:25:13
Speaker
openly they're going to be made available but if it comes personal information it has to have a consent of the individual so for example a bank making a check on their land record credit score etc so that's the consent message that goes okay okay they are informed that hey you are going to get a message because of this please keep on that okay got it okay so what is this product called the the product which is throwing up the credit score It's called Satsource and it's actually part of a larger product suite for the bank, which is called Satsource Sage.
00:25:48
Speaker
Okay. What else is there besides this credit scoring, which we just went through? So the other one is the village ranking ah module, which provides the leads that the first example that you are asking me. And the third one is the ah monitoring ah piece, which is is called SatCollect, the the con continuous monitoring of the performance of the crops across the portfolio. and all the, let's say, ah changes, anomalies that we are detecting because of drought, stress, flood, all kinds of climatic events that happens, ah creating a signal for potential default or delinquency for the banks so that they can prioritize their farmer connect accordingly.
00:26:35
Speaker
Okay. So, I mean, once you've built the data stack, then these are all different use cases for the same data side because the monitoring product is taking the same data. It's the same ability to detect what is being grown.
00:26:48
Speaker
ah you're able to detect not just what is being grown, but also whether it's healthy or not, like when you're saying. Yeah, it stresses, it's healthy or not. And there are, it's a complicated kind of event detection engine where news articles let's say pest or disease attack, prices fell, something like that triggers immediate anomaly detection algorithm on the images. And then we you know ah run like this is like beyond the scheduled monitoring.
00:27:21
Speaker
So what like the color changes to show you that this is pest attacked? Or like how does that 40 pixel tell you that this is pest attacked? Yeah, ah but Kestradagon will not be just in 40 pixels. It will be like, across the village for the same crops.
00:27:37
Speaker
Right. And yes, it is through the color changes, which happens because the photosynthesis starts reducing and the photosynthesis reduction and reduces the chlorophyll content content in it which it, which changes the color. That chlorophyll, that is what is captured by satellites, because a satellite image that you see on a Google map is a color image. It's a RGB color image. What we are analyzing is ah ah is not just that. It's beyond into narrower bands of infrared ah spectrum, ah which provides this kind of signals that are not visible necessarily in the color bands.
00:28:18
Speaker
OK. OK, let me zoom out now. ah what when you say earth observation what all data is collected so you said rgb you said infrared uh can you just first talk more about what is earth observation what all data is being collected how is it being collected uh are these satellites same as say the tesla uh sorry the spacex satellites which are providing internet are the similar kind of satellites but just they have a camera on it or like you know just For a complete layman, tell me what is Earth observation and how the Earth is being observed?
00:28:53
Speaker
ah That's a great question. I think we should have started with this. yeah So, so yeah let's take the zoom out. um the The ones that you mentioned, like Starlink, that's for providing internet connectivity. Those are communication satellites. They're different kinds of satellites. uh what we are speaking of is earth observation and there is third kind of satellite which is for navigation like the gps satellites are ah for that so these are the three different kinds of uh services out of uh satellites that are created communication navigation earth observation within earth observation uh it is basically uh uh you know
00:29:33
Speaker
dip is basically depending on the discriminating factor of ah reflectance ah of different land features like you know, soil has a different reflectance, vegetation has different reflectance. Within vegetation, trees have different than, you know, ah crops. Within crops, at different crops have different kinds of reflectance values taken over a period of time, not necessarily one single time instance, that is used to ah identify ah and quantify ah the changes that are happening on on the planet. Now, this is... One second, sorry.
00:30:09
Speaker
What does reflectance mean? Right. So... ah Within the Earth Observation, there are ah primarily two types of sensing technologies. since One is optical and multispectral. The second one is the the but radars which is active microwave satellites. I explain what is sensing in that.
00:30:32
Speaker
When you what you see from optical multispectral satellite in that camera, there a sensor like any other camera. So in that sensor captures the the light that is reflected from the surface of the earth.
00:30:47
Speaker
because of the sun, you are getting more reflection from the earth. And that reflection reflection passes through different filters, ah filters of different bands of NIR and all, you know, the basic physics that we were studying in ninth. And that creates the image ah within the camera, just like any other camera here. And this is called what, ah um you know,
00:31:12
Speaker
passive remote sensing with passive remote sensing is basically optical and multispectral satellites hyperspectral is a niche uh so you mentioned pixel so pixel is a peer company a good friend of ours partner of ours actually uh they are on uh uh they they are basically making ah these satellites which have very narrow bands ah like instead of ah a let's say 10 band which is a multispectral satellite okay they will have 200 300 bands which are much narrower and are able to even go down to a property of a soil and identify potentially minerals etc
00:31:53
Speaker
ah So it's a specialized form of optical multispectral satellite. The other one is the act active remote sensing, which is a radar. which is the radar So operate ah in the microwave part of the electromagnetic spectrum. So optical multispectral is basically your Vibgear RGB and then the towards the red part of the spectrum. Okay. Thermal infrared, near infrared, etc. On the other end of the electromagnetic spectrum ah is a microwave.
00:32:23
Speaker
Beyond that is UV rays and all. So my this this operates in microwave part of spectrum, but this is an active sensor because it doesn't need ah the sun reflectance of the Earth from the sat ah from because of the sun rays. Here it is sending an active pulse.
00:32:40
Speaker
Okay, as a radar, it is sending an active pulse to the Earth's surface while it is moving at 7 km per second and receiving it back. And based on the... path based on the characterization of that pulse as it comes back, it is able to tell you the features that it's looking. So the difference here is that ah when even if there is no even if there is no light, like at night or with huge clouds, you will still get a good sense of what's happening with SAR. In optical context, optical multispetal clouds will mean that you will have to attempt to capture the same image in the next attempt.
00:33:19
Speaker
What is SAR? Synthetic aperture radar. Sorry. Okay. Okay. So it's a type of radar. Okay. yeah so so the the The radar one is relatively easier to understand. It works just like any radar works. It pings and the reflect, the ping reflects back. And based on that, it's able to see land features. Okay. So I understand that. So is the land, like whatever, is there a building there? Is there a room there? Is there a tower there? or So those things it would be what the radar one would do. Um,
00:33:48
Speaker
multispectral, hyperspectral, ah these are still, ah I'm still struggling a bit to understand these, to be honest. i don't know if you can simplify these a little further. Yeah. So ah very difficult to do that without visual size failure. If you look at an image visual, it becomes easier because if you recall the ah plus two physics stuff, that's the electromagnetic spectrum. you know at And I'm a commerce student, so that makes it doubly hard for me. ah ah Yes, yes. But, ah you know, it's basic, it's just how your camera works. ah You take your phone camera or DSLR or whatever. So, at night, at in dark, your camera will not collect any information, right? Because it's not seeing anything.
00:34:34
Speaker
yeah Exactly the same. Take this camera up its face, sensor, mirror. Optical multispectral is one single thing or these are two separate things? optical can be It can be same or it can be separate also, meaning that in a camera, you can put only the optical part, which is the color images, RGB, etc. RGB, that's it. Multispectral would be this near infrared, short wave infrared.
00:34:57
Speaker
Now, there are big satellites that can accommodate all of these filters in one single telescope, like the camera taking here. Whereas smaller satellites, they can only go with RGB, just the optical. But the principles of working is same, which is why I categorize that under the same, you know, passive. it. Got it. Okay. Okay. Multispectral is just a better sensor on your camera, essentially. That's it. And hyperspectral is like ah even better. Okay. Okay. Okay. Got it. These comes with the trade-offs. That's the only thing.
00:35:25
Speaker
So what is the additional information that you're getting through infrared?
00:35:33
Speaker
So in infrared, um the reflectance of, let's say, ah I mean, there there are different bands within infrared also. The reflectance prop ah of water, soil, vegetation, ah these are very nuanced.
00:35:53
Speaker
uh against in color uh images for example you can uh kind of quantify the changes in moisture of your canopy and your soil uh through this uh kind of infrared data okay yeah so it's not just 40 pixels of ah rgb on the basis of which you are telling which crop he's growing but it's all of this additional information which you're getting from infrared ultraviolet which is helping you to uh accurately identify what is the crop, what is the health of the crop, things like that.
00:36:27
Speaker
Absolutely. Because within that 40 pixels in one single timestamp, you will have 13 different you know band information so it's like a stack okay now you keep imaging the same place again and again with 13 13 bands so you have a very dense information very dense information okay so this is like a massive amount of data which you need to crunch store and crunch and like i'm assuming your aws bill must be pretty high ah Yeah, I mean, we are technically a very big data player. Also, ah very important and key partner for AWS in the region.
00:37:14
Speaker
Okay, okay, interesting. Okay, okay. So, ah ah now, you've told me how, like, the the camera works within the satellite, but I want to go a little more zoomed out. What are the...
00:37:26
Speaker
I believe even in satellites, there's like low Earth orbit and stuff like that. Are there different types of satellites, different types of orbits? water What are the satellites which are collecting this data? Who's running them? Why is this available for free? You know, how is it being monetized? Like someone, I mean, there's no such thing as free, right? Someone is paying for it. So I want to understand that a little bit more. Who's operating these satellites and so on?
00:37:48
Speaker
and right so not all satellite data is free uh the only some of the public space agency satellite data will be free uh which uh is a good amount of data now who is paying for it of course taxpayers why is should it be free because it is a digital public good we should you know uh be able to build ah such kind of applications that has a uh economic and social impact on uh on the country right yeah so uh the the position to make such data free was actually taken initially by usgs ah us government uh and followed by uh european union who devised this multi-decade long program uh make the uh commercial use cases research uh like you know accelerated research on this kind of data uh
00:38:39
Speaker
as a larger i would say long-term vision of uh uh adapting to climate change driving sustainability you know innovations uh and and solving you know real life problems okay okay okay got it okay um there's like say gps that which helps us navigate that's free right like and that is uh so similar to that so that's also like a public infrastructure a digital public infrastructure so similar to that this data is uh by these government ah space agencies is provided for free absolutely but just like in gps what you get is a civilian signal with a five meter xy like positional accuracy uh whereas the military signal of sub meter positional accuracy is is
00:39:27
Speaker
you know secure it's only with the U.S ah government so same as the case with earth imaging satellites uh low resolution satellite like this 10 meter pixel even isro has a 5.8 meter pixel image equal resource site it's free now uh but very high resolution images like what you see on Google Maps for example those are produced by commercial companies like Maxar, Airbus. So you will see attribution even on Google Maps that these are contributed by Google. So they pay. Google Maps essentially you know pays these satellite operators to provide what they call as base map.
00:40:01
Speaker
OK, OK, OK. So what or what is the orbit? What are the type of satellites which are doing this? So these are essentially launched by, say, a European Space Agency or an ISRO or a NASA or ah like the Japanese Space Agency. they are organizing these ah satellite They are launching these satellites and maintaining them. And the data is beaming back to them. And then they are uploading this data on the cloud, making it available for research use.
00:40:30
Speaker
Yeah. So the Earth imaging satellites, which we are using for our purposes, these are all low-Earth orbit satellites. ah The ones that we use, commercial and space agency included, would be anywhere between 400 to 700 kilometers from the surface of the Earth. ah The GPS ones, on the contrary, are in medium Earth orbit, which is 8,000 kilometers from the surface of the Earth.
00:40:59
Speaker
The weather satellites that we are ah having, ah those are in geostationary orbit, which is 36,000 kilometers above the surface of the So these are the three most ah utilized orbits. There are some specialized orbits as well.
00:41:14
Speaker
okay so what are each of these orbits for like what's the reason why earth observation is on low earth orbit obviously because your camera cannot capture from 36 000 kilometers right but but just can you like again for a layman just break down each of these three layers of low earth orbit medium earth orbit and geostationary orbit what do each of these mean Yeah, so look low earth orbit is basically it's been ah historically the orbits for imaging satellite because otherwise you will have to build huge telescopes to be able to see the earth at the ah you know kind of information density you get. ah So that's a engineering trade-off that okay I'm gonna send my satellites ah up to maybe 500 kilometers, my satellite size is 200 kg and it will have a life of maybe eight years. if
00:42:12
Speaker
ah If, let's say, you had to do the same specification of image capture, let's say one meter resolution at 36,000 kilometers orbit, the size of that satellite will explode so much that the... Yeah, it will be like Hubble or James Webb telescope or something like this. Yeah. Okay. The use case defines the ah orbits. ah For example, a GPS, it's an 8000 kilometer orbit because the functioning of your GPS is to give you your location and the movement velocity ah ah with which you are going. Right. So for that to happen, if you are ah you know putting these satellites in low Earth orbit, so it has a much smaller spread. Think of it like this is Earth's surface. You come here, Earth orbit.
00:43:03
Speaker
This is like only small area. You take it 8000, it has much bigger area than it is over. You take it to 36000 kilometers. then your signal strength will only reduce. Then you don't have the necessary latency with which you get that. So there are so many factors of the use cases that defines this um what we call as the mission.
00:43:24
Speaker
okay Mission specification. And what is the deostationary orbit for? Well, traditionally that has been used for communication satellites and weather satellites. ah And that's also continuous. But the the the communication satellite has... So communication means like the Starlink internet? No, no. So that's what I was explaining. Communication historically has been broadcasting DTH, that kind of stuff.
00:43:52
Speaker
Okay. Okay. And even there was like mobility on the sea, in the air, ah for for that purpose. However, Starlink changed the ah game here because they prioritized internet. ah Now, internet from 36,000 kilometers came with a higher latency. ah But obviously, ah to bring the satellites down to low Earth orbit and provide internet connectivity required more number of satellites. Now you can contextualize why Elon Musk is sending 10,000 plus satellites there. okay
00:44:23
Speaker
But he could because he owned the rocket. So hence, these satellites, Starlink satellite, OneWeb satellites, which are meant for internet connectivity, are also in low Earth orbit.
00:44:35
Speaker
OK. Because you need low latency, whereas for a broadcast, there is nothing which is being sent back as such. It's just one-way signal. And even if there is some latency of a few seconds, it's not a big deal. got it That will not work on it in internet connectivity. OK. Understood. ah What is the number of Earth observation satellites which are currently out there?
00:44:58
Speaker
Yeah, I don't know. if Maybe 1,200 or so. Because there are that's it okay yeah there are military ah satellites also. something but Small number compared to like 10,000 of Starlink that Elon Musk wants to do. okay I'm feeling like I'm some space rapper on chat GPT right now.
00:45:18
Speaker
okay But 1200 is a good enough number for whatever earth observation data that one needs.
00:45:27
Speaker
I guess, yeah. I mean, it's it's it's it's it's not good enough per se because ah there are several data gaps ah that happens, you know, um out of the 1200 a huge portion of it is military satellites so those are not commercially available uh so yeah if you bring that context it's not enough what you have in terms of earth imaging satellites like if you have to take a geopolitical context around it China would have about 300 satellites which are doing the earth imaging and India has what eight ah government I'm speaking of government satellites so if you have to
00:46:09
Speaker
kind of look at benchmarking that is the asset deficit wow okay okay interesting uh and does China make the data available do you have access to that China also makes some uh uh the low resolution their uh satellites uh data available So you said there are some gaps in the data. What are those gaps? What is it that you is on your wish list?
00:46:35
Speaker
What you currently don't have? It's all right now coming up. For example, hyperspectral data ah ah that that is not commercially so rampant. Indeed, the Pixel has been a part breaker ah in in you know building their satellite constellation, um followed by, I would say, Y1, which is a Canadian company. Then there are... ah um data gaps in core thermal infrared what we are using is near infrared short-frame thermal infrared companies like satview hydrosat uh constant ir these are companies which are building it so um yeah they they are fairly new companies young companies so maybe a few years from now and that gap also will be filled by such companies and how will this thermal data help you in better decisioning for your clients
00:47:23
Speaker
Oh, lot of things. You can identify urban heat islands, you know, and monitor them. ah And then you can do so many things. ah I mean, just as a very vanilla flavored example for for ah ah relatability, you can... drive ah advertisements of ah buying ACs. Okay. Okay. Okay. Or you could tell quick comments like Zepto and Blinkit that talk more ice cream here.
00:47:59
Speaker
ah okay Okay. Fascinating. Okay. So we've covered one product of agricultural use case. What are your other products? Right. So for us, so ah we have always looked at our business from a land feature ah that we can identify using the images. So one is so at a top level, there is vegetation. Within vegetation, you have the crop and the trees. So I explained couple of you know use cases. Crop related.
00:48:32
Speaker
I'll explain on the tree part because when we look at an image, we are not just looking at crops. We are looking at trees, roads, buildings, everything. So the best utilization of your image and your infrastructure is if you have models that can analyze all of these features ah and you package and you know provide solutions to different customers from the same data using the same mod, similar models and same but software platform infrastructure. So from a tree's perspective, there are multiple things. one First is ah ah in providing trustworthy data ecosystem for governments um who are
00:49:09
Speaker
and know spending money on aforestation programs because they have you know raised external capital also ah from international funders like World Bank, Jai Gai, etc. So they need to also show that how the capital utilization has been, what's the impact metrics, etc. So ah this becomes a very important tool ah for them ah to not just demonstrate the effectiveness of the program and its implementation, but also to ah have continued deadlines ah for for and you know kind
00:49:42
Speaker
keeping this these programs running second is uh in tackling deforestation uh which is an important aspect uh you have human uh wildlife human nature conflict are going on everywhere and uh the the ability to act fast uh seeing the deforestation triggers ah is something that every government today prioritizes third is uh how do you also help governments uh to kind of connect these natural assets as part of their land here to ah you know monetize in carbon markets because trees are one of the biggest sources of sequestering carbon into the ah from the ah atmosphere. so
00:50:28
Speaker
uh identifying what is the carbon stock uh and how it's moving etc help helps the governments to build or engage into product projects uh that are globally acceptable to uh you know play in the liquid like carbon uh bonds and carbon credits carbon credits market now this is one part of the for forestry part uh one one quick second sorry um Many people may not be aware of this carbon credit thing, but essentially like there are ah bodies which will pay dollars for carbon reduction. And the way to claim those dollars is by showing proof that you have reduced carbon and either by growing trees or whatever. And so this is the market that we are speaking about here, like the carbon credit market.
00:51:15
Speaker
Am I correct in my summary? or Yeah. That's that's the... Okay. Got it. take it got it Okay. So for this carbon credit market plus fighting deforestation plus ah project ah monitoring for aforestation projects, so you actually have government clients or you're talking of this as a possible line of business?
00:51:39
Speaker
no no we have government we we have been working with government of Rajasthan in India for more than a year now uh we have uh uh other like state government clients uh that we are working with I mean at least the government rice and portal is uh public there's YouTube video the others I don't know if uh yeah but yeah okay but it's not just in India we are working outside of India as well we are working in Australia across forestry and I agree Okay. Okay. So in Australia, it would probably be to fight deforestation. Like you would send an alert that somebody is cutting a tree here. like like That's what the alert is, that someone's cutting trees here?
00:52:18
Speaker
And if that is legal deforestation or not. Legal meaning that they have to get permission to clear as some part of the forest to do something else. Okay. Okay. Okay. So you also ingest the data of who's taken permissions. That permissioning data is available with you. So you're able to flag this. And it's our Australian partner.
00:52:33
Speaker
ah okay that's with the Australian partner got it okay okay okay and okay understood so this is your second product uh about trees okay uh yeah please carry on yeah but in trees there are other things also that are important for example there's a compliance angle to it uh now we export a lot of different commodities agri commodities like coffee tea uh you know soya uh all of these uh require the uh sustainability certification uh to be exported at the right price to markets like europe japan us etc and uh uh
00:53:16
Speaker
There are international now international nos to ah show proof that ah the produce has not been sourced from farms that were previously forested.
00:53:29
Speaker
Okay. So, ah the only way to do that is satellite images, by the And it is also to be done within certain like plot level. So, again, the same ah land record, ingestion, etc. that we have anyway done comes into play here also. ah And this is not just for the producers and traders.
00:53:50
Speaker
There is compliance angle for also the product. FMCG companies who are finally sourcing it like the Unilever, Nestle of the world and the banks who are funding them in terms of ESG reporting. So ESG reporting ah ah requires to prove that they have not funded deforestation in their portfolio. So yeah,
00:54:14
Speaker
so yeah I mean, look, the the thing is... Here your your paying customer would be the FMCG company, right? Not these individual ah farmers. yeah Yeah, so you know we have plethora of customers across. FMG companies are customers. Banks are customers for this.
00:54:31
Speaker
Okay, for the ESG. But in India, there's no ESG focus with banks, right? Like these will be global banks. Yeah, these are global banks. Okay, okay, okay. core And again, just for my listeners, essentially ESG is environmental ah social governance sustainability governance sustainability and governance sorry okay environment sustainability governance essentially in the west if you are esg company which means you are doing good things for environment sustainability governance then you get loans at a lower rate of interest and which is why then banks need to validate and make sure that you're not just doing something which is known as greenwashing but you're genuinely ah working on ESG okay understood okay so yeah i understood this product also have we covered your portfolio of products or are there other use cases
00:55:24
Speaker
We will never be able to cover that in an hour. But yeah, it' it so it's kind of adjustments, data, adjustments, because the same ability of us to ah you know detect these trees, type of trees, species of trees. Like, for example, we work with FMCG companies on tree species identification also because they do the sourcing of pulpwood, for example, eucalyptus, etc. You know, we have to do also hides of trees there. And this leads us directly into ah power line industry. Why? Because power lines, they have to, you know, route, ah they have assets that are going through forested areas. And there are high chances of wildfire and, you know, any kind of storm that leads to the trees falling on top of the ah the the wires, which again, is service disruption, a lot of money. So ah even transmission companies are our customers because they need to do the right-of-way management. The right-of-way is essentially the clearance ah within the wooded areas through which the transmission lines goes.
00:56:27
Speaker
Okay. Fascinating. Okay. ah What else in terms of product lines? ah Well, you know we have ah ah we have ah ah and like aviation product ah and insurance product here.
00:56:43
Speaker
ah But yeah, I mean, these are again not the industries seem like so a sudden change, but the capabilities on the technology stack remains the same. If you are deleting heights of trees, you remove the tree, make it a building, we can detect the heights of buildings also, we can detect the building footprints also, right? So ah that goes into your airport perimeter monitoring, the changes across the airport perimeter, which leads to changes in the flight path, charts, procedures, all of that digitized in a single platform. So as I said, we are again, we are not selling just intelligence, translating that into outcomes. So ah an airport operator and the their survey department, cartographic department don't have to keep on now, you know, manually making all of that within like four months.
00:57:33
Speaker
it is auto-generated and validated on our platform. And are you working with Indian operators or global also? Both. So we work with AirPot Authority of India. they have been They have been an anchor tenant client for us since 2021. And yeah, we also work now outside of India on aviation.
00:57:56
Speaker
OK. And insurance? Same here. We have both Indian and... What is the insurance product like? Just tell me the product first. I mean, whatever I told you, buildings, trees, etc. what Crops, when it gets damaged, you have to do loss adjustment, remote assessment of how much is the damage. So, this is where your before-after analysis becomes very critical. That's what we provide.
00:58:20
Speaker
This for crop insurance specifically or like what kind of insurance? Even buildings, not buildings roofs, ah roof caved in, earthquake. Okay. What is the level of damage? ah What are the loss adjustment against some insured Those are the things, scal of decisions that are made based on the image intelligence data that is provided.
00:58:38
Speaker
Okay. So instead of sending a surveyor on field, this is all done through ah earth observation. Yeah, because the customer experience, ah ah if an insurance company can say within 10 days, I can clear your claim versus a 90 day process because the adjuster could not come to your place. So obviously people want to simplify it for the final consumer. Hence,
00:59:03
Speaker
These are the digital, let's say, trends that leads to satellite earth intelligence becoming a silent back-end partner for all such initiatives. Okay. You might be aware Lenscard bought out this ah company called GeoIQ.
00:59:20
Speaker
ah Yes, the geolocation intelligence company. So what was that for? That was for like similar, their their output is similar? Like telling Lenscard what's the best place to set up a store?
00:59:33
Speaker
Yes, but that's based on, ah so basically you source public data ah like population, socioeconomic data, you generate some data or yourself on top of it and then you bound them into, let's say, pin codes, areas, etc. And then you say that, okay, based on my mechanism of rating disease factors, this is the best area. So that does not include this kind of computer vision applied to satellite images. Okay, that's not Earth observation. that It's just the other data which is layered one on top of another. Yeah, location. Your data.
01:00:09
Speaker
Yeah. Okay, okay, okay. What are the big untapped opportunities which you haven't even touched so far? Well, there are millions of it. To be honest, I don't know. Low-hanging, like which are the low-hanging fruits that you... Low-hanging fruit. Like you haven't mentioned much in terms of retail or marketing, know, in those spaces.
01:00:31
Speaker
Yeah, I don't think ah there is an enormous value that companies like us can add in those spaces. Why is that? See, intelligence from satellite ah doesn't compete ah with the ah in with with with any other sensor like the drone versus satellite that ah argument itself is stupid. ah So we compete with
01:01:05
Speaker
human intelligence okay as evenly can a person do this faster cheaper and uh at lower and ah with with high reliability with consistency which means uh any kind of intelligence that you draw over large area assets whether it is you know forest crops or roads or transmission all of these things huge span uh there it becomes a slam dunk wherever it's like city retail intelligence for city you send people you will get so much more richer data because satellite will give you one dimension in the rural areas e rural of it becomes very difficult and costly okay so it's it's costly to cover large ah land area which is where ah eo helps but if it's a dense if if there's high density it's much cheaper to just have a guy on a bike roam the area and give you that local level intelligence and
01:02:02
Speaker
Okay, okay, okay. Very interesting. Okay, so ah what made you want to quit ISRO? Like, take me through your career path and your journey a little bit. Like, how did you end up joining ISRO? What did you do there? so it's been some time, 12 years. I quit ISRO because I wanted to do my further studies.
01:02:25
Speaker
So, yeah, that was the trigger point for me. Because otherwise I'd have to wait a lot of time. Since ah ah but within an organization, there are people who are more senior to you who may be waiting also. So, ah yeah, that was a personal call.
01:02:42
Speaker
And yeah, I ended up going to France and even doing my master's. That was what triggered And then after you did your master's, which year was this when you did your master's? 2013 to 2014. Okay. Then what did you do after that?
01:02:56
Speaker
okay then what did you do after that Like what led to the birth of Satshaw? Take me through that journey. Satshaw was born in 2017. In between 2014 and 2017, I was working with a boutique research and consultancy firm for space sector called Northern Sky Research. And yeah, that's where... That was also in Earth Observation, like similar...
01:03:20
Speaker
it was ah like overall all communication, earth observation, everything, manufacturing, like everything related to space sector consultancy. So um yeah, it's a Boston based company. And I was initially an analyst within the earth observation, and small satellites segment of practice, we should say. um So yeah, i got the opportunity to um work with the interesting entrepreneurs, VCs, large companies who already you know are in the business and mostly senior level, which kind of helped me also understand the nuances, the big picture of the industry. I think that was quite instrumental in having the, let's say, first form of the vision of why SatShore should exist.
01:04:14
Speaker
Like, why should it exist? but What was the gap that you saw?
01:04:19
Speaker
Well, The main gap was so the structure of the market itself that it was more supply centric and not demand centric. People were launching satellite without telling precisely what problems are they solving. Because they know that if they launch satellites, which is an engineering, like it's a difficult engineering problem to do, ah whatever be the case, defense will anyway buy something out of it.
01:04:42
Speaker
ah Because they are always hungry for more data. So ah all the capabilities was, ah you know, directed without specific outcomes ah and especially the kind of outcomes that I thought are more relevant relevant to our part of the world, the global south. um I think some the companies also were starting to do well like Planet which is now a listed company, they did very well. um and that gave us confidence that this is possible it's feasible we just have to and it is again this uh at the intersection of earth observation and AI yeah I mean they have been primarily uh on the earth observation side they operate the largest suite of flat lights they're San Francisco based company and now they own the asset also so they're like a hardware plus software kind of a company
01:05:36
Speaker
So they own the hardware and they are now building the analytics layer. Like that that's the main focus for them now, which is exactly opposite of Satsure, where we have built the analytics and software layer. And now we are doing backward integration and launching our satellites.
01:05:54
Speaker
Okay. Okay. So i'll I'll revisit this backward integration part. But yeah, okay. So you saw that, ah essentially, you saw that people are launching satellites. And like, if you build it, they will come kind of an up approach that if you create satellites, use cases will come up.
01:06:11
Speaker
And you saw that there are a lot of use cases. Now, we've just gone through a whole bunch of them where humans can be replaced through AI plus Earth observation. And that was the trigger point to start Satshore.
01:06:24
Speaker
yeah and what was the ah
01:06:31
Speaker
like you essentially started your the journey with the banks only like that was the first customer that you built for no actually we started our journey with government okay the forest uh with agriculture only okay what what was that first uh like the first project you did The first project we did was ah in a district called Shrikakulam in Andhra Pradesh.
01:06:56
Speaker
It was a chicken and egg problem for us that if we go to a commercial company, they will ask who has validated it. If we go to a government, they will say who is using it already. So among the two of them, ah the government customer at that time seemed an easier one to crack because, you know, ah agriculture is primarily a state subject and there is always intent to do more for for the farmers. So um I happened to ah kind of meet ah through a good friend of mine, Mr. Ram Mohanadu, who was a member of Parliament of Shrikakulam.
01:07:36
Speaker
uh he's today the union minister of civil aviation oh wow okay and uh he was at that time serving his first I think tenure as an MP and he was just a year older to me it's very young very dynamic so something he saw and he said yeah okay what do you need from me and like okay if you do a pilot project with in your district we need support from he wrote a letter to the district collector that to help these boys they are trying to do something interesting it works it's good and it will look good for the state and for the farmers so that letter was what we went to the DC and he was also quite curious he ah enabled us to visit farmers visit the agriculture university scientists so that we can be guided well and I think that that project as we completed ah we learned a lot
01:08:21
Speaker
ah That gave us ah the opportunity to also go and pitch for a larger ah work with Andhra Pradesh government and through a ah summit, ah a challenge called AP Act Tech Summit in 2017 November, which was backed by Gates Foundation, Willem-Lyanda Gates Foundation. We also got the opportunity to pitch there and we came, like we won the challenge with this international jury selecting our approach to solving that problem. and A couple of months from there, we had the the government of Andhra Pradesh under leadership of Chand Babu Nadu almost seed funding us with a 5 crore purchase order.
01:09:02
Speaker
ah What was the problem you were solving for the government? It was ah essentially a digital crop monitoring system where we would build all of these capabilities of vision-based crop identification, identifying the sowing progression, harvest progression, because we realized that some of the decisions, et cetera, requires these rate indicators, like what's happening throughout the cycle. So that helps them to give contingency plans through the Krishi Mechanic Centers. Harvesting means that's a signal for them to you know provide the services to farmers. etc and all of that. So ah we mapped out ah clearly our capability development requirement to serve the banks and insurers, which are crop detection and yield and expanded it to meet the government's ask.
01:09:50
Speaker
Okay. Okay. Amazing. You bootstrapped this for quite a while, right? So these ah government projects were what were paying the bill? Yeah, yeah. So ah we bootstrapped for all well four four and a half years. Yeah.
01:10:04
Speaker
And was it err voluntary or you were forced to bootstrap because nobody was funding? let' see in my previous job uh i had uh interfaced a lot with both vcs and bankers uh uh this is circa 2017-18 uh peak uh you know fomo on fintech and e-comm and then edtech during covid so you know We are fighting for the same pool of capital, right? And we are, there is so much complexity to explain to a VC. So, like, why to waste time and do this? You know, just ah be capital efficient, keep costs low, you know, get customers to pay for it at at least get to a thesis of how will you show exponential growth?
01:10:52
Speaker
Like all companies need not go to VCs. It's only when you need to grow fast, you need that capital. So 21, when you raised your first round, you reached a certain thesis that, okay, this is the path to growth and this is why I need funds. What what what what were the answers that why do you need funds and how will you grow?
01:11:09
Speaker
Because we had to launch the banking project in 2022. ah Okay. So you could clearly see that there's a large market. What is the size of the market that you are catering to?
01:11:21
Speaker
like I'm sure you would have put some TAMs and all those numbers. I'll say some number which is based on bottoms of calculation. It's okay. It's like that's a big TAM. Just think about it. Like a cred, for example, ah built a business around credit card users, which would be about one and a half crore of Indians. Now you imagine now how many farmers are there in the country and total services. So, yeah.
01:11:49
Speaker
so it's a be much much bigger time but a slightly unproven market urban niche users are of course better bets this is a national bet this is it's not just money but also potential to create impacts of technology okay okay your founding team is all isro folks like you have two other founders right one or two founders yes so ah how did uh how did you build that uh uh commercial culture uh because you know coming from isro i'm sure it's not in the dna to think ah customer first product first revenue first uh how did you solve those problems of becoming financially viable as a business uh when you bootstrap for uh you know
01:12:35
Speaker
yes And survive. This is survival instinct. Yeah, yeah, yeah. i trade Every business, ah you know, lies between your top line and bottom line. So, at the end of the day, what we do at Statshore also helps people either make more money or save money.
01:12:53
Speaker
oh How do you price it? Like, is there a, like, it's a, every case is priced individually or is there a, like a science on how you're pricing it or is there some thesis you found?
01:13:04
Speaker
like advice on how to price? It is very much a custom pricing for people. So we don't disclose the pricing. of Because we go to the depths of what business value are we actually delivering here.
01:13:21
Speaker
And it's a fraction of it. Like how many manors you're saving? Essentially, that's the, yeah in most cases, you're saving them a lot of manors. Exactly. Okay. ah So, you know, I think from what I understand, your business is not exactly like a full self-service productized business, but it's a mix of consulting and product, right? Because I'm sure every time a new customer comes in, there is a lot of consulting initially before it becomes a product.
01:13:50
Speaker
Absolutely. ah Not every place. ah ah With the banks, it's fully productized. It's like one single API people can integrate. Yeah, that is highly repeatable. Like people coming for loans. But like say the forest ah thing, is that also totally product or is that like consulting?
01:14:07
Speaker
So the thing is, ah ah the the notion that people have when we say product is that there should be a nice web UI, SaaS subscription, etc. To us, the product is everything that happens behind that, because that's where the entire repeatable reusable all data models, etc. comes in, right? So that one fine layer of application for the end user is the consulting part.
01:14:31
Speaker
Got it. Okay. So the product is for your internal team to use, essentially. Your internal team can very quickly answer any question ah a customer comes to you with by using the product, which is all of the set of capabilities which you have built out.
01:14:46
Speaker
Right. And when we create create a thin client layer on top of it understanding what they want, how they want to consume this data, and then they also get access to the same data sets that you know we have access to, maybe, yeah, limited.
01:14:59
Speaker
What is the kind of team you've built? Meaning size or? I mean, like these people who interface with clients, are they data scientists or are they business analysts? Like what? what and And, you know, so so just from that structure, like what are the different... ah teams within the organization who's doing what so uh interfacing with clients are done by uh people who are uh you know who have lived a technical life before uh and have have adopted a salesy uh hat so these are not like mba folks who are like okay these are uh
01:15:42
Speaker
mostly tech folks, IITs, my college, IST, who have kind of donned a more technology consultative action and the sales team and them work together.
01:15:54
Speaker
So have you heard the term called forward deployed team? No. What does that mean? So it became a very commonly used term nowadays because of the AI wave. um So forward deployed engineers are those guys who you know do the necessary plumbing with a customer ah to get the model, et cetera, architected within their existing ecosystem. So they are folks who can write their own codes, have deep domain expertise, and act like you know savvy business consultants. so
01:16:27
Speaker
uh this this entire concept was actually invented by a company called pallet we started in 2003 now it's a hundred billion dollar listed company what we do in our context we don't need the software engineers engineers we have the you know forward deployed team which is a combination of your product manager your data scientist your business analyst and your sales team who who anchors them and ah they do the cracking How did you build sales capability? Because I mean, as a founder, you can go get a few doors open, but how do you scale it up beyond yourself?
01:17:02
Speaker
So first and foremost is ah um I learned the hard way that unless a product ah has has truly achieved ah PMF, ah it's important that the founder only does the sales. So once it is, ah you know,
01:17:22
Speaker
achieved, which is where I think the signal can be a bit fuzzy, kids considering we are shaping markets and monetizing at the same time, um then you can bring in a sales team because you have a proper SOP that this is my ah you know customer persona. This is that list of similar customers. This is how you sell. you So ah having that certainty on the process, the prop the product and the ah customer journey is when you build a sales team.
01:17:51
Speaker
And for that, it's okay. Anyone who selling to banks, fintech solutions can come and sell this also. As long as they don't have to come and answer that in 40 pixels, how it come in? Because you have gone through that. Right, right, that's right, right.
01:18:04
Speaker
Okay, okay, okay. Fascinating. um You know, how ah how supportive is the VC ecosystem today for deep tech? Like, you have now, I think, till date raised more than $15 million, right? So...
01:18:20
Speaker
We have raised about $25 million. $25 million. Wow. Okay. what What were the various rounds, various stages? and think your first round was a $5 million round, if I'm not mistaken. ah Right, right. And then ah ah we did a 15 million round.
01:18:34
Speaker
ah Then there is some 5 million we raised through both grant as well as existing our investors. only So okay this to me, see, we we started working way before deep tech was a thesis, ah we way before space tech was hot in India. ah And ah hence, most of the current VCs with deep tech thesis didn't have one when we were looking to raise money. So we were lucky to have a a few good investors backers at an early stage. So very private equity. India is our lead investor. and the
01:19:18
Speaker
you know it was very interesting because their entire history is coming from private large private equity public market investments uh indeed i think they still hold more than one percent hdfc uh so uh them doing a smaller fund early stage thing was interesting why because uh they didn't come with a thesis that you know as an early stage investor i'm going to be ag tech or fintech or know climate tech or whatever their fundamental question was yeah Who is buying your stuff? Why are they buying? We want to understand that.
01:19:50
Speaker
we will frame our thesis. How big this is. And the other thing is, how did you sell? How are you executing? We want to understand that. and and basic fundamentals of ah business.
01:20:03
Speaker
So they underwrote, I think, the ah market potential of what we do. They underwrote the founders ah and the rest, I think they allowed us the freedom to ah ah run our experiments on both the GTM and tech so that l we continue going forward.
01:20:23
Speaker
Okay, okay. This 15 million round that you did is because now you said you are doing backward integration and you are actually owning satellites now. So this is to fund that.
01:20:36
Speaker
Yeah, we are not owning the satellites yet. So yes, this round was in order to build the capability to make the hardware piece as well. ah And of course, also refine our go to market internationally on the you know AI analytics piece. So ah now we will be raising more capital. ah to fund the commissioning of the assets through the public-private partnership initiative by InSpace that our consortium won. So, Pixel was the lead, SatShare is the SatShare and ZorovaSpace and Pearsight are the other partners in that.
01:21:15
Speaker
What is this InSpace thing? Can you just zoom in a bit on that? It's first of its kind public-private partnership to have a privately built and operated constellation of Earth observation satellites that you know will serve the domestic country needs as well as built in a way that it's internationally also basically they support to create an international Earth observation company out of India. so by uh coming together uh pixel sasha adhruva we are side uh we essentially ah you know created a very unique single company or spv here which has all types of sensors our hyperspectral optical venti sector everything
01:22:00
Speaker
and the algorithm layer and the because that that layer and the ground stations etc everything in one place at droop space is the ground stations yes okay are you speaking to them yeah I've interviewed Sanjay yeah but that was like two years back so it's been a while oh yeah so yeah you didn't ask all the questions on Leo Mio zeroo to him two years ago so I forgot Plus, somebody who will be listening to this will be in that frame where they wouldn't have heard a two-year-old episode. So, yeah, I just wanted to make it easy to access.
01:22:39
Speaker
Okay. i so So, you formed an s SPV and so this the government is funding this, you said, ah or it's partly government funding, partly private funding?
01:22:50
Speaker
It's primarily ah ah private funded. The government had offered to part fund it ah here, but we are doing it ourselves because ah we it was anyway part of our business plan. Pixel is not making their constellation. We are making ours. taking government money at that time didn't seem like a great idea because ah you know this is the first time initiative we didn't we want them to move fast as well what what is the public here in this public private partnership what is the government doing
01:23:24
Speaker
ah government is acting as the uh anchor for it in terms of review uh reviewing the progress uh providing the technical expertise working uh on getting buyback commitments uh because even now about 200 250 300 crores i think annually civilian different departments uh uh you know they buy data from foreign companies our sources buy data from foreign companies which is totaling to some some other 400 500 crores so if your captive market with with very little awareness already is 800 crore data market huh what can you make it if you bring everyone together data platform solutions
01:24:05
Speaker
So, okay, I guess that time question, I didn't actually pin you down on specific numbers. But you're saying that the various ah parts of the government of India are already spending close to 800 crores buying ah earth observation data and intelligence.
01:24:23
Speaker
Yes, between defense requirements across the force different forces as well as civilian, it is that much. Wow. So globally, this is like India would obviously be a very small spender. Globally, this would be a pretty massive spend that is already there. What what is that number? The TAM number?
01:24:41
Speaker
ah Globally, so it is very big. because a US defense alone consumes, I think, $4 billion dollars of imagery a year. And I'm not even speaking of solutions on top of it. um And if you look at it, how much governments spend on space, especially space defense, it's quite high.
01:24:59
Speaker
I think recently Germany constituted a $35 billion dollars space defense fund. It's all over the news. France has like a space fund of only space technology fund of $5 billion dollars and $100 billion for ai on top of that.
01:25:15
Speaker
Wow. Okay. Okay. so So this is a massive market, essentially, this market for Earth observation data, which already exists. Now, ah this... ah ah kind Your ah hardware plays essentially through this s SPV, this in-space ah s SPV that you're doing. It's not outside of that.
01:25:35
Speaker
No, it is ah so it is in-house as well. For example, um we spent our last few years, three years in developing... novel ah payloads. Payloads is the camera system that goes to space. And while we are you know providing the same payload to the s SPV here to you know build and launch the multispectral optical satellites, ah We are also having a line of business where we partner with companies like Dhruva and others to provide our payload to them because they manufacture the satellite, AID, all of that they do. And that happens from our subsidiary Collideo.
01:26:20
Speaker
Okay. So this is essentially the camera equipment that you're doing. Yes. And this is built to order. This is not like you're putting up your own cameras, which you will operate, but you're selling it to satellite ah companies who are launching satellites and you will possibly buy the data from them or something like that.
01:26:40
Speaker
No, no. We are simply at for them providing the license of our IP and it's a bill to print contract. Meaning that they you know they can get in manufacture or they can get the contract to us. We'll get in manufacture and deliver the system to them.
01:26:55
Speaker
Okay, so you're selling the IP. Okay, got it. Got it. Got it. Because ah for us, the need to do the backward integration was to have data sovereignty and have levers on the pricing of the data to keep opening up the commercial market. If you can't control the cost of it, how will you control the price of it? So the cost of it is only when you have designed that system to meet your the demand you are catering to at a very different unit economics than what you know people are making in the US and Europe which is for the same quality of data which is much higher uh I I I'm not fully understanding this can you just like explain to me one more time so you're saying you need data sovereignty what does that mean so if tomorrow uh again uh let's say uh political relationship between uh uh US and India uh dips right now also it's not great uh
01:27:53
Speaker
The different levers that people use to put pressure are the following. Trade. That's one that comes in the news. Second are the strategic items. Slow down the image access to ah companies. and Yeah, like US stopped sharing intelligence data with Ukraine after the disastrous Zelensky meeting.
01:28:13
Speaker
ah Okay. So that's a business continuity risk for a company like us. So and some alternatives so so you're saying that you need... satellites uh indian satellites which have the kind of camera sensors ah giving you the data which you need for business continuity and therefore you decided to build camera sensors yes we decided to build only after checking you know whether the others can build it for us um
01:28:45
Speaker
the The thing is the market is skewed towards the defense demand, ah which is not the kind of only demand we are catering to. We are catering to commercial demand as well, which requires a lot more wide area imaging and all. So if it has to be built from scratch, so we will be have we will be paying for the entire IP development for someone.
01:29:10
Speaker
Okay. We are buying off the shelf then. yeah It was off the shelf, you would bought it and simply flowed the satellite. So because it is IP development, there's no sense to do it outside. We can do it inside. ah you know ah From the background that we came in, this is actually the easiest stuff to do than building the products for banks insurance and all. Right. Right. Right. so what is so This, that IP that you have created, you're not actually having your own sensors out in the sky.
01:29:44
Speaker
You are just selling this IP to other companies. But by having a lot of companies use, probably to these companies, you're telling them that if you launch satellites with this kind of sensors in it, I will buy the data from you.
01:30:00
Speaker
That also, ah but we will have our own satellites through the SPV, right? Through the SPV also. Okay. yeah okay Okay. But that may not be enough to cater to all the demand. So hence, yes, ah by selling our IP here, this is like an option value. We also diversify the data acquisition ah places. yeah You're able to tell companies that launch a satellite with my sensors. I will buy the data from you. You have to pay for the sensors, but I will pay you for the data.
01:30:30
Speaker
That's it. ah okay okay so currently how much of your data ingestion is paid and how much is through open source uh right now we are like having maybe 40 of our data needs are commercial 60 are open okay okay and these commercial how much is through global companies how much through your indian partners or most of it is global companies All of it is Google companies because our Indian partners, ah for example, hyperspectral data, yes, pixel. But there is no other company, Indian company, which is providing high resolution 0.5 or 0.3 meter images.
01:31:11
Speaker
So that is currently a gap in India's space strategy. Like we don't have that those kind of satellites out in the sky. Yes. Which in space is going to solve now. in spaces PPP is going to solve with Ministrins, we are going to solve. Right, right. Yes. OK, amazing, amazing. OK, let me end over here. I think I have got most almost all my curiosity satisfied. um I just want to like kind of end with some advice which you may want to offer to builders who want to do deep tech startups. What what kind of mistakes have you made which you'd want them to avoid? Or is there any advice that you'd like to share with future deep tech founders?
01:31:52
Speaker
well you know uh i think in deep tech uh any kind of deep tech domain uh the most important thing is to keep your costs low uh tap into all kinds of development grants for non-recurring engineering costs that you need uh and that support is only increasing i mean look at the rdi one lakh crore uh so these These are ah some of the important things like survivability is important because you have to underwrite the overall development or commitment for Decade Plus. This is not a, you know, spray and pray kind of domain that you try multiple things, move fast, etc. Sometimes it's okay to be slow because your market is still evolving. um
01:32:39
Speaker
other is uh i think uh just just being careful in uh how the same thing is communicated again and again with your colleagues so that they also know what they are getting into and they don't get into let's say uh phases of burnout and disillusion etc so it requires ah a lot more uh i would say um closer collaboration and communication to execute properly because most people- you need a mission-driven team, basically. And you need to build that sense of mission drive in your team.
01:33:20
Speaker
Absolutely. Okay. Okay. Okay. Fascinating. ah One last question I want to slip in. ah this ah With the way the ah the models, by say like say Google's Gemini model, et cetera, how they're progressing,
01:33:36
Speaker
ah Do you think that these models would very soon be just able to take that satellite data and provide the kind of insights which you have spent a couple of years training and developing?
01:33:51
Speaker
so um There will be certain tasks that ah will be done very well by ah you know some of these big tech models. Already, I do see ah basic stuff like the building footprint, etc. being quite good I mean it's evolved very fast uh but what it cannot do uh still ah is uh that context addition and further you know modeling and productization right so um like throwing out a credit score is still beyond the capability of a model Yeah, even to just identify ah your and my firm next to each other half of acre, that will not will
Automation in Business Processes
01:34:35
Speaker
not do it. It requires very different training. interface So ah it will it is actually an accelerant for of companies like us because we can automate a lot of the back end other tasks using these tools. So it's not a threat. It's actually a very good movement.
01:34:51
Speaker
Okay. Okay. Amazing. Thank you so much for your time, Prateep. I took way more time than what I had on your calendar. But I was just so curious about what you're building. Thank you.
01:35:04
Speaker
It's my pleasure. Thank you, Akshay. All the best.