Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Service Provider Executive Series Episode #15 - Weaving the Future of Regulatory Innovation with Ari Caroline
51 Plays1 month ago
In todays episode, host Steve Gens sits down with Ari Caroline, Co-Founder and Chief Strategy Officer of Weave, to explore how generative AI and platform thinking are reshaping the regulatory landscape. Ari shares Weave’s origin story, the metaphor behind its name, and how the company is building a “superhighway” for regulatory content generation.
If you'd like to get to know Weave more, please visit their website at: https://www.weave.bio/
Recommended
Transcript
Ari Caroline's Transition and Background
00:00:07
Speaker
Welcome to the Gens & Associates Regulatory Executive Podcast Series, where we explore innovation in the regulatory space. I'm very excited about today's guest, Ari Caroline, who is one of Weave's co-founders and who recently transitioned from founding CEO to chief strategy officer.
00:00:24
Speaker
Ari, before you introduce yourself and weave, I've been in this space, I don't wanna say it, 35 years. I used to be a Johnson & Johnson person for a while before I went into consulting.
00:00:36
Speaker
And you see twists in terms of different innovation at different points of time. And I mean, a lot has happened in the last 20 years, but it's unreal over the last two to three years and we'll get into it.
00:00:50
Speaker
i think it's still early days. It's just a really credible and a really exciting time. So what you and the team at Weave does, I'm sure our listeners are going to be really curious to learn more. So before we get started, can you introduce yourself and Weave to our listeners? And you know certainly Weave has a very interesting history. So maybe you want to share that for our listeners too.
00:01:10
Speaker
Yeah, I'm happy to. Thanks for the introduction.
Weave's Vision for Data Consolidation
00:01:12
Speaker
So yeah, we spent probably the first six months of the founding of Weave thinking about what this new paradigm is going to look like. And I think the way you described it is correct. It is in a truly new paradigm that is emerging that is being empowered by the technology.
00:01:29
Speaker
And you have to think about where that paradigm leads us in terms of the ecosystem as a whole in order to realize, you know, in the order to find your place, what you and your company can uniquely contribute.
00:01:43
Speaker
So the original inspiration, while I was still at Tempest Labs, now Tempest AI, they had bought our first company, was that there was so much in terms of technology and data sets that was just dispersed, disparate in the drug development ecosystem that it it just didn't make any sense, right? These things contribute one on top of the other, and you really need a platform ecosystem that brings it all together.
00:02:13
Speaker
And as you know as we discussed before, it's not too many steps from there to realize that once you've gone past the drug discovery process and you're in the ah process of drug development, everything that happens around drug development is around regulatory.
Regulatory Evolution and AI's Role
00:02:30
Speaker
And so if you're building a backbone, if you're building that sort of foundation for a platform like this, it makes sense to build it around the regulatory process. Thanks so much for the introduction. And as you were going through it, I had a picture, I'm a very visual person, of a dossier, right?
00:02:49
Speaker
standard thing that happens every day, you know thousands of them them go to the health authorities, but if you would sit back And especially with a new product or even CTA-IND, if you would kind of sit back and map like the different chunks of content and more of the listings, the data tables, you know, is it from 10 source systems? Is it 100? Is it 250? And it's this, I think most people that are not in the space are not in regulatory.
00:03:17
Speaker
how all this comes together. And even though there's been great strides in technology in the last 20 years, some of it's still very manual. And I think that's where we start because probably 25 years ago, had the ECTD and introduction that led to the concept of regional regulatory affairs, but also It's about a standard, right?
00:03:38
Speaker
So there was process change. There was organizational strategy change about regional hubs. And it took a while. So people got very serious about document management. We have our style guides. We have our templates.
00:03:52
Speaker
And that's been pretty static for a long time, you know? Structured content authoring, kind of up and down, always trying to solve that for labeling, probably more applications in the clinical space. But with generative AI and certainly some of the things you guys are doing, it's like all these points are starting to come together. And I have never seen in my career when I was in industry or now,
00:04:15
Speaker
the potential of rapid transformation and just the whole, and I don't even think we should say the authoring paradigm, it's content generation, right? And how can we more effectively for the first or second filing, but also lifecycle management? So maybe it's a great opportunity with this rapid transformation. How do you guys play? And I'm more interested how you see it, you know, like in the short term and where is this train headed?
Integrating Data in Drug Development
00:04:39
Speaker
Well, so the name of the company was deliberate. We chose the name Weave because we saw this as the opportunity to weave a tapestry that works, you know, that continues throughout the process. I don't know if you've ever seen the Bayou Tapestry in Normandy, but it tells a story of the Norman Conquest.
00:04:57
Speaker
That was a very long tapestry that's laid out over many different walls. And it's telling this narrative that's building up over time. And it's, in some ways, very similar to what we do in the drug development process. And when you do these updates and the amendments, right, you're doing them on, you know, material and previous studies that have been done earlier.
00:05:17
Speaker
And it does need to be all connected. But we didn't have... that sort of universal view, that continuous view into the dossier previously.
00:05:29
Speaker
It was all scattered in these separate documents. And in terms of sources, I would guess that, you know, when you're talking about the entire regulatory lifecycle, it's thousands ah of different sources that you're bringing together.
00:05:40
Speaker
And in terms of chunks of information that are being repeated and updated throughout the process, you know, you're talking about combinatorics that get much greater than that. Right. So you need to have it all connected did in order for you to be able to search it, in order for you to be able to update it, etc.
00:05:57
Speaker
It was why when like we made the commitment to help with the authoring process and you said, it's really content generation, it's not just authoring. The authoring is just one piece, right? Then there's all the refinement and review, the QC process, the submission, and then that's a single submission.
00:06:14
Speaker
But that's part of a much larger dossier, right? And so that all needs to be one continuous process. It's why we made the decision not to build our capability as a simple word add-in.
00:06:27
Speaker
Because that creates the focus on the single, first of all, creates the focus on just the authoring process, but also forces, narrows your view to that single submission or that single section.
00:06:40
Speaker
And our goal was always to create to have this in terms of the larger view of the dossier, of the platform, of the life cycle of the drug. So I want to follow up on one point and I might be taking a tangent here. So let's see, because when I say recently, the last year, there's people that are in different camps that it's The future of structured content authoring, even though that's been kind of up and down, more prevalent in the med tech space, more application in the med tech space.
00:07:11
Speaker
But also we're in the early days of generative AI and there's also a school of thought. It's not one or the other. The better structure you can bring in any content, the more you can leverage AI. But Like, how do you couple that? And maybe for our listeners, like, where's weave taken that? i I love the metaphor about the tapestry and all that.
00:07:30
Speaker
So you're weaving, it's about information connectivity, because what we see in our research in our regulatory benchmark data is And it's pretty dramatic in the last two years, instead of classic system to system and connectors, we call them the DAPs, data aggregation platforms, data lake, data warehouse, data mesh. There's just so many names to those.
00:07:52
Speaker
But the whole content generation. So if you're weaving, especially mid-tier and larger companies are doing a lot of investments. So you hear concepts about authoring our content generation over these dApps, you know, maybe a few years out. I just thought our listeners to get, you know, more a little bit into the architecture about literally how do you weave it all together?
00:08:13
Speaker
Yeah. So I think starting with the point about structured content authoring, think the goals for structured content authoring were always, you know, right on point. That is the goal for where we want to be at the end of this process is like where everything, you know, there's a table that you can reference at any given point. And if you want to update it, it's updating it consistently across all ah across the entire dossier.
00:08:38
Speaker
The challenge is that there are so many different variables that we need to think about, and they're very different from one submission to another, from one program to another.
00:08:49
Speaker
So I was part of a sideline discussion at DIA with a group of folks that are working on the FDA prism in that initiative, that Veda Perkins, and we are contributing some of the AI capabilities there.
00:09:02
Speaker
And we had identified this one use case around M4Q in CMC submission, right? And that was a standard that was created just a lot of time and effort was put around it, but really focused on small molecules.
00:09:17
Speaker
And what you need to know about stability of a small molecule is very different than what you need to know about it an antibody, about of m mRNA cell therapy, right?
00:09:28
Speaker
And it's not natural for us to be able to think about every single feature that you would want to manually enter into a process like this. But what the AI technology gives us the opportunity to do is not, I think if most people have worked with the current AI technologies would say that they're not very good at working with structured data.
00:09:50
Speaker
But they actually can offer a translation layer in which we can extract all of that structured information. And then since we have that structure, the document structure, the causal structure in the background of our systems, we allow you to draw on it repeatedly, getting consistent responses.
00:10:10
Speaker
So almost like a mapping. It's like mapping algorithms. That's right. And then you can think of the LLMs, the AI, as the translation layer between the source documents and the underlying systems. And then as you're drawing upon that underlying data representation, putting it back into narrative form.
00:10:30
Speaker
Yeah, I'm just amazed how the ah just the technology, the architecture of the technology. I was at a conference earlier this year and this brilliant person had just joined this new organization. i don't know if they were like one of the lead innovators at Google. I forget the company, but it was a fun conversation where if you look over the last year,
00:10:52
Speaker
Oh, let's take a guess at it, 40 years. There's so many technologies. Some come and go, there's fads. I'm big believer of the Gartner hype curve. You can look through all those things and eventually, you know, it makes it out the other side, just like the drug discovery. There's a lot of trial and error and some of them make it and a lot of them don't, but what's fundamental? And so we said, well, if you could only choose three, mine were, and actually ours were almost identical. So our consensus was the personal computer.
00:11:22
Speaker
the internet and now AI. And these are like tectonic shifts. So a little sidebar there. i don't know if you would agree with that or maybe add on, but- Oh, I would i would totally agree.
Technological Shifts and Their Impact
00:11:33
Speaker
And I think maybe even to take it a step further is that when we have these tectonic shift type technology movements,
00:11:43
Speaker
What we see is not just an ability to do things faster. There's a great quote from the physicist Philip Anderson in and an essay that he wrote, I think it was in 1972, where he said more is different, right? When something happens at a scale that surpasses a certain boundary, everything changes.
00:12:04
Speaker
The paradigm changes. You have emergent behavior that is just very different from what you've seen before. Anything that's true with faster is different as well, I would say. So going back to an older technology, if you think about the advent of automobiles, right, with shift from the horse and buggy to this engine that can take you, people say, oh, it goes a little bit faster. So now you get to your destination a little bit faster.
00:12:28
Speaker
But what you actually saw emerge in terms of the ecosystem was a whole different world. You know, these roads got built that inter interconnected everything, every city, every small town, every house.
00:12:40
Speaker
You had... You have new communities that emerge because you can have people living in suburbs and able to commute. And I would say similar with the internet now, right? We're both having this conversation from our home offices without the level of bandwidth, without that level of speed.
00:12:57
Speaker
Something like this doesn't happen. We can't function this way. So we start to create an entire world around that new capability. And I think that's what the AI represents here. And it's very much the way we're thinking about the technology in building the Weave platform is what are we empowering?
00:13:14
Speaker
Are we just making a faster car? I think the closer analogy for the way we see ourselves in this ecosystem is we're building one of those super highways that's going to empower many other technologies so to come on top of it.
00:13:29
Speaker
That's a perfect segue into my next question, because as you know, there's so much focus and rightfully so for new product development, time to market, time to patient, and everybody's working hard to reduce that. And it's just amazing from when I was in industry to 10 years ago,
00:13:46
Speaker
to some of the data that we saw last couple of months that last patient database locked to a filing best in class according to McKinsey. It's like 10.5 weeks, which is kind of unheard of. And people are looking at single digits.
00:14:00
Speaker
But that's where I know you guys play in that space and support that process. But where I want to go to is, and any regulatory professional listening and knows that 70% on average, depends upon the company, of regulatory activities and cost and resources goes into lifecycle management. but You have your obligations, maybe there's CMC, manufacturing change, change control process, label change, et cetera.
00:14:25
Speaker
So could you, as far as what you do and how you contribute at Weave to not only the new development process, but it doesn't get talked about enough, the whole lifecycle management and that complex complexity, if we think about it mathematically, the permutations there are just unbelievable.
00:14:42
Speaker
So what could you share with our listeners on that? I think it fits well into that faster is different paradigm.
Supporting Drug Lifecycle Management
00:14:49
Speaker
So just to catch you up but in terms of the Weave platform as it exists now, we made a deliberate decision to start at the very beginning of the process.
00:14:58
Speaker
So that's like pre-IND briefings, the IND enabling studies. Through the IND, earlier this year, we released our first capabilities for clinical stage submissions, including CSRs.
00:15:11
Speaker
um We're but going to be able to do amendments, response to health authority questions is the next module that's getting released in just a month from now. Going all the way, including protocol authoring, every single part of the clinical process.
00:15:24
Speaker
By the end of this year, we'll be supporting all the way through approval. So we called the first module there auto-IND, and then auto-CT is the clinical stage submissions. Now it's all part of the same platform, but reaching all the way to auto-NDA, auto-BLA for the approval process.
00:15:42
Speaker
But that's by the end of this year. The goal for next year is to move into post-market and think about labels, pharmacovigilance, et cetera. But getting back to what I said before about faster being different,
00:15:55
Speaker
you could think about now the opportunity if bringing a new drug to market or a new variation on an existing candidate or an existing approved drug, right, is, as you said, just another permutation, right? And if the process for going through that, either at the development stage or afterwards in post-market, is a much simpler, faster, cheaper process, then our ability to customize these products offerings, actually, it starts to become realistic. not just I'm not just talking about rare diseases.
00:16:29
Speaker
You know, like we know that a lot of these drugs are successful in some classes of patients and not in others. And sometimes it's just a formulation, right? And if you can customize it to the metabolism of a certain class of patients, and you can do many more permutations on that, that enables Many, many more capabilities as well.
00:16:50
Speaker
So absolutely, we plan to address this as well, not just because it's the 50 to 70% of what actually is submitted, but because it's all connected and all and it's all about enabling that ecosystem as a whole.
00:17:03
Speaker
What I would emphasize, though, is that our decision to start at the beginning was deliberate. Because even post-market, when you're making these decisions, if you want to just modify a formulation slightly, right, you're building upon everything that's been studied before.
00:17:19
Speaker
And you need that broader picture. for what happened from the beginning to the end. And that goes into this notion of having this long tapestry that is your dossier that you're continuously building upon.
00:17:32
Speaker
That's telling a story and and building upon a story that's been told until now. And it also sounds, and um I think they call it the Merlin factor, where you start with the end in mind. So as you're going down the path that if you start with the end in mind, it just think about all the reuse and the referencing.
00:17:51
Speaker
And I'm thinking about Weave now as, you know, Weave connects all the dots, right? a little different from the tapestry. So that's fantastic. I think that gets into the last question that we have. And I usually like to end on this and especially it's so fascinating in the space that you play. it's It's so new, it's going so quickly, it's so dynamic, it's very different, but just the potential benefits and still you and others are in the early days. I just couldn't imagine what it's going to be like in two or three years. It's changing so quickly.
00:18:24
Speaker
So from a short term and a longer term, what should we expect from the weave team?
Future Plans and Ecosystem Development
00:18:29
Speaker
Well, so in terms of the short term, it's very straightforward, which is covering that that entire regulatory lifecycle, right?
00:18:37
Speaker
So the HAQ response health authority questions capability is a very new one ah for us that required us to build new technology. But it builds upon everything we've had until now in terms of having all of your data available in a centralized data room that's an AI-native data room and being able to draw upon that And then bring those questions in and find the appropriate sources for those responses.
00:19:02
Speaker
A lot of what comes next is the same thing. It's building upon everything that we've done until now. And so that's the approval stage was built on that same ECDD structure as you get to your NDAs and BLAs.
00:19:15
Speaker
And as we discussed, post-market it is also important. It's a continuous building process. And so we want to cover that entire regulatory lifecycle by the end of this year through approval and by 2026 having a full post-market offering as well.
00:19:32
Speaker
That's the path that's laid out that highway. You know, where if we're building that central highway, we know what direction it's going in. But the branching points from that highway, as you said, are more than we might be able to imagine.
00:19:46
Speaker
And that comes down to being respectful of how little we know about a paradigm shift technology like this one and where it can go. And so our approach instead of over-optimizing for the current capabilities of the current AI capabilities, we've always been looking at the limitations as well and saying, how do we get past these limitations?
00:20:11
Speaker
So it used to be that putting together really good structured tables was very difficult to do with the early AI capabilities, that they're getting better and better at that, but you still need to know how to massage it and how to structure it.
00:20:24
Speaker
And so that was a capability that that we had to build. I think in terms of structured offerings, structured content, and being able to extract that structured content, present it in useful ways, is another edge of that of the technology now that we're making a lot of inroads in being able to do follow consistent computational and calculation processes that follow the guidance.
00:20:47
Speaker
And that can automatically update based on as the guidance gets updated, um where you actually update your content, your code for how you calculate something like m MRSD based on lots of animal tax studies.
00:21:02
Speaker
So we have to stay at that leading edge of the technology. And that's we've dedicated this moving into the next stage of the company, a percentage of the budget that goes into this we've researched space that is really focused on staying at that bleeding edge.
00:21:20
Speaker
ah But I think we also need to think about not just the off-ramps of this highway, but the on-ramps as well. And so if we're really building something that's meant to be a central highway, how do we empower lots of other things to build upon the ecosystem that we've built as well?
00:21:35
Speaker
Going back to that original inspiration of the company of really needing a platform that allows you to stitch together all of these capabilities, where you don't somebody creates a new platform Uh, tax prediction model, right?
00:21:49
Speaker
The current paradigm or the past paradigm has been, you've got to figure out how to market that, turn it into a product and sell it and break your way into a really tough industry. What if there was a paradigm in which there is a platform that, you know, you can just build that on top of? And if you've got something that's built around the regulatory process, it's very natural to create an app that that's a tax prediction app that's built on top of that core platform.
00:22:17
Speaker
And so thinking a little bit further out, that's one of our goals as well. Yeah, and I think the other shift, it's implied with what you're saying is the previous era and the spotlight, if you will, is we call them transactional systems, your core different
Enhancing Regulatory Efficiency
00:22:34
Speaker
systems. But on, as you're saying, this new super information highway,
00:22:39
Speaker
It's about connecting the dots, connecting pieces, not so much systems, but pieces of data to accelerate the process. but It's very different. It's very exciting. It's probably, and to that point, we're talking about two sides of a platform of a system here.
00:22:54
Speaker
Right. There are the sponsors that are preparing of the submissions and then there's the health authorities, the regulators that are actually reviewing it. And it's very obvious that like you want consistent technology that allows these two parts of the system to speak to each other seamlessly and make sure that whatever story that you're telling as a sponsor is the one that that, you know, the health authorities are receiving in their review process.
00:23:23
Speaker
Yeah. So what I learned today about Weave is it brings the concept of end-to-end to a whole new level, you know, versus within function to how you really kind of laid out from the beginning and how you're evolving.
Closing and Future Reflections
00:23:35
Speaker
As an organization, what I'd like to do is some of our listeners and might want to get hold of you. So what's the best way? Is it a website, LinkedIn? What would you suggest if somebody wants to have a question or follow up with you?
00:23:48
Speaker
It's really easy to reach out to us through the website and everybody in senior leadership gets any inquiries that that come through that website. People who want to reach out to me personally, I'm at ari.weave.bio.
00:24:01
Speaker
um So that's easy to remember, ari.weave.bio. We love to hear ideas and thoughts in terms of where the platform is going and ways to partner with lots of different folks that are doing lots of interesting different things in the space.
00:24:17
Speaker
Yeah, thanks so much for that. And for our listeners, you know how to get a hold of us, go through our website or I'm a LinkedIn junkie, as a lot of people know, just reach out on LinkedIn. So thank you again so much. And we'll have to have you a back because this train, it's a speed train. It's a bullet train, you know, going down the tracks and it'd be great to check in with you. And just, I think,
00:24:40
Speaker
organizations like we've thought leaders like yourself and your team, that that is the future to where we're going. So thanks so much for spending the time with us and sharing where you're at, your thoughts with our listeners.
00:24:52
Speaker
Thanks for inviting me, Steve. And the yeah, we're excited to share the updates as we have them. And there are going to be lots. All right. Sounds great. Thank you so much.