Alex Mok: Building Biotechs in the AI Era

Ram (00:02.007)

Hey Alex, welcome.

Alex Mok (00:03.906)

Thank you. Thanks, Ram Thanks for having me on the podcast. This is a conversation I've been looking forward to.

Ram (00:09.848)

Yeah, myself as well. As a quick introduction to our audience, Alex is a founder, ex-founder of a biotech called Mantra Bio. was the founder CEO and currently at Nveda. And it brings a very interesting perspective on biotech and the journey which he took from starting early and sort of progressing through the process of building his company.

as well as what he's doing in his career right now. I actually noticed in your LinkedIn, you say you position yourself as a zero to one guy. So I'm actually quite excited to talk to you about that journey you took and your learnings there,

Alex Mok (00:55.554)

Yeah, yeah, excited to talk about it.

Ram (00:58.894)

All right. So I think we should start with maybe your journey into the biotech space itself, right? Clearly you've gone through the ideation and let's go found a company, let's raise some capital. So that whole very early stage of the process, could you tell us about what was that like and how did you end up there?

Alex Mok (01:22.136)

Yeah, yeah, it was a bit secluded. I would say I didn't plan to go into drug development, you know, after postgraduate education, but effectively my, my first company, which I was on the founding team was developing a liver organoid drug, add me talks platform. So organoids were starting to get a lot, a lot of excitement today.

This is a company called Sally sick, which we had co-founded in 2025 Sorry, 202 that 2005 20 years ago. Geez So this this was a long journey before you know organoids sort of hit the mainstream but effectively the platform that we were building was a way to evaluate drug toxicity within artificial liver organoids and the way those are constructed were instead of running a

your typical tox studies in animals, actually running them in fully formed organoids that were developed from human cells. And in that process, I actually got a very keen view into the drug development process. Although on that side, I was not the drug developer, nor the one creating the NCDs or what have you. It was mostly creating platforms, as well as research tools in order to enable drug development.

And so through that journey, you know, worked with a lot of different companies, pharma and biotech on how to identify various talk signatures and the like. That company ended up getting acquired by EMD Millipore and Merck KGAA. And then I was sort of on the inside the umbrella of a very large pharma, Merck KGAA, and basically started to continue to build platforms in that journey.

and started to get really excited about, the potential for platforms that could develop drugs. and, from that experience, as well as a few others, came around to starting my second company, which was mantra bio, a exosome therapeutics company. And this was in the, the space of drug development, sorry, drug delivery and drug development, in order to identify, you know, homing vehicles that one could pack.

Alex Mok (03:46.6)

RNA, small molecules or genetic information to specific cells and tissues in order to generate drugs. So that's actually kind of how I got into the drug development. I was sort of adjacent or orthogonal to it and then jumped right in once I saw a platform kind of concept that I could get behind and could see kind of the road towards building drugs. now I'm managing a number of programs and also involved in a number of

different clinical programs here at INDATA as well.

Ram (04:19.764)

And could you tell us a little bit about what did it take to really get from an idea to starting up a company in the biotech space from your own journey?

Alex Mok (04:32.888)

Yeah, yeah. Well, each journey was different, but I'll focus on mantra. think mantra, our thesis was that exosomes, are a, you know, nature's effectively drug delivery vehicle, or you could say nature's communicator, right? Cells secrete vesicles, almost like nanoparticles, very abundantly, and use those to actually transfer.

about material information, proteins, et cetera. And our idea was, can you use cell engineering techniques, state of the art cell engineering techniques, as well as AI to basically build the perfect engineered drug delivery vehicle. And so this was a step towards a new modality, which is a very humbling process, especially in therapeutics. We hear about small molecules, oral tablets, we think about.

injectables like monoclonal antibodies, you those, those technologies are, are quite mature. But when we're thinking about gene therapy, cell therapies, exosomes included, you know, you really have to reinvent every little piece of drug development from, know, how you manufacture it, what the analytical assays, what's the PK assay, what's the talks, you know, that you can run because all the interactions with kind of the biological space is a little different.

And in mantra, think our biggest challenge was building all of that scaffolding for us to actually make a drug that we could give to patients. So when we started to think about clinical development in which therapeutics, we wanted to develop what was a target product profile, what was the indication, was the patient subset that we were targeting. That's when

you know, technology and platform and development really became more of a preclinical into clinical development, which is a totally different view of what I think most technologists think. because then the, you know, as we all know, the road to developing a drug is 10 years. I think the latest number is $2.4 billion or something around there, between two and 3 billion per drug to get approved. And that amount of planning capital outlay, you know,

Alex Mok (06:53.96)

Derisking that you need to have at various junctures is a very, you know, long, humbling process that I think takes a lot of rigorous decision making. So I hope I answered your question, that's sort of my experience.

Ram (07:09.646)

Yeah, yeah. And that's an interesting thought process, right? Because if you're building in any other sector outside of biotech, know, your goal as an entrepreneur would likely be how soon can I get to some kind of a profitability or, you know, some kind of evaluation which will let you sort of stay alive default. And I think what you just said about taking a 10 to 12 year journey,

and value and get in some manner early on and potentially signing up to raise about two and a half billion dollars. If that's what it costs, that's probably where it takes to kind of get through all the way, right? Now, of course, there's inflection points where you get value and you get the opportunity to get acquired and things like that. But I am curious in the really early parts of your fundraising journey, how do you sort of go about and say, this is the value of this asset or this idea I have?

And I know since you mentioned it's a exome based technology, is it more of a, it's not a specific drug for a specific therapeutic area, right? This is more of a delivery mechanism or a platform to potentially help, you know, make other drugs better through the new modality, that right?

Alex Mok (08:26.646)

Yeah, our vision at Mantra was to build a platform which would then develop drugs from the technology of that platform. So in particular, one example is we were quite interested in ophthalmology and specifically targeting the retina. And so we did a lot of research and development around exosomes that could be targeted towards retinal cells, RP specifically. And the idea there was not that we would just generate one retina drug from that, but that we could generate many.

right? You know, anchor them with different proteins, anchoring them with different, you know, cargo payloads. And, you know, we really needed to build both the pipeline and the platform at the same time, which is a really, you know, I think, arduous effort. Very different than a single asset, you know, drug play, where you, you know, quite focused on just generating one asset, the investment, the capex that you need to generate the platform.

which includes manufacturing and pipeline is, you know, you have to think about it as kind of competing costs.

Ram (09:34.638)

Okay. And again, going back to the way you value and how do you sort of like get started, what are some of the factors which maybe your initial investors took into account when they were saying, okay, is a bet we need to make,

Alex Mok (09:51.66)

Yeah. Well, so drug delivery has been a holy grail area for decades. You know, there's been a lot of interest in different drug delivery vehicles, such as LNPs, liquid nanoparticles, AAVs that we all hear about, other modalities as well. You know, you could throw ADCs into there as a form of drug delivery as well. And what we were trying to basically stand up was a completely new nature-inspired

drug delivery platform. These are vesicles that forms a communication layer that exists abundantly, as I said, across cells and tissues. And can we just identify those sort of routes, so to speak, and then use engineering in order to build on top of them and exploit those routes. And when we kind of began our early pitches with investors, it was really focused around this point. Can we generate basically a

you know, generational opportunity based on trying to solve this holy grail drug delivery. Um, and I think the timing was very important. Uh, when we started the company in 2017, there was a lot of new modalities, RNA, uh, you know, mRNA, was well know of, uh, si RNA as well. Uh, you know, gene editing, Cas9, CRISPR, those things were starting to take form. And really the unlock for all of those modalities is can you actually get them to the right cell?

so that they can make their genetic modulation or manipulation. And to this day, we still have that. It's still a holy grail. And if you ask any investor, what would a company that cracks that code that becomes basically the category leader for drug delivery, what's the valuation? I think everybody would say north of 50 billion. So it's a really hard problem to get the right drug to the right place for the right patient.

Even saying the word hard is a massive, massive understatement. And I think that that's sort of how we try to attribute value is showing that each inflection we had on the technology development actually led to a completely new unlocking capability that would be very unique for the company and be a value driver.

Ram (12:15.852)

And because you sort of are talking about how there's a capability to do a lot, right? This is a platform company. You have different therapeutic strategies. I'm curious from a capital allocation perspective, does that mean then you say, well, I have this technology which could potentially be used to drug various different targets and I will go after a certain route. And I'm just curious, like, how do you pick that route? Why RP?

Alex Mok (12:39.789)

Mm-hmm.

Ram (12:45.166)

versus something in oncology, right? And what are the decision-making paths which you probably took before you said, this is where we need to go or we should go?

Alex Mok (12:55.734)

Yeah, yeah, that's an excellent question and very involved, I think, in terms of the answer and still one where I'm pinning down an answer. But I think there are three prevailing wins, let's say. One is obviously biological and science. Where does the biology guide you towards? The tissue example of the eye. We did a lot of experiments to understand where we could actually see an outsized distribution.

Ram (12:59.918)

Nothing really, yeah.

Alex Mok (13:24.6)

through different administration routes. And so we tested systemic through IV, we tested intraocular, we tested many other routes, intraperitoneal. And just saying, okay, where do these exosomes go? And if we know generally where they might distribute, can we actually target them even further when we add certain ligands to them? And so that was kind of the first step in terms of.

Where does the biology teach us? What can we learn about these, this natural pathway that we can exploit? And the second is market driven. So 2017 to 2021 is a very different market than 22 to 24, which is 25 and 25 to 28 is going to be a different market too. So I think, you know, uh, I think most people wouldn't disagree with me to say that 2017 to 2021 was a very platform centric biotech market.

There was a lot of interest in platforms. think Moderna is a bit of a poster child of that one as as many other names that you've probably heard of. But the whole thesis of platforms is we can generate a ubiquity of drugs with the technology, And kind of Stefan Benzell is famous for talking about mRNA as software that I can just kind of program and say, have a new RNA, I a new molecule, therefore I have a new drug. so investors love hearing that because if you have,

ROI continuing in perpetuity off of a platform because you can make drugs and each drug is worth you know, anything from a billion to You 20 billion then that's that gets a lot of people quite excited as as as kind of Capital markets shift and of course biotech in particular given the large runway that you need to get to approval Very very sensitive to where the macros are, you know where inflation rates are etc You start to see a shift in 22 23 24

to assets, you know, farmer are buying companies for assets or they're not so much buying company for platforms. Yes, you see that, in terms of the value, it's not even close. We're talking about a magnitude difference. So at least so, so I think at the time that we were raising, you know, the story was around platforms and that was something that was credible. That was something where the market was valuing and we were there were many comps comparables, you know.

Alex Mok (15:44.409)

Today, would we do the same story? Probably not. We would have to really build a story of, think, getting to the clinic and de-risking the asset in patients as soon as possible, the various catalysts that you need there. But I think in 2018, 2019, where companies were IPO-ing upon submission of IND into the clinic, it's a different piece. So biology, market, and I think lastly, obviously, is what the

the culture, the vision of the company. We wanted to be a therapeutics company, but we also wanted to have kind of a hybrid partnering strategy as well. Because if we, the bet is we can achieve ubiquity, why can't we do both? I think in, obviously there's a lot I could probably share about that strategy and how that played out for us. But I think that those are the types of questions that we thought about. I think many other founders in the biotech world.

think about when they're starting the company.

Ram (16:46.168)

Yeah, no, that's excellent. And there's so much to unpack in what you just said. In fact, my own journey, right? I think it overlaps with those years you're talking about, 2016 to 2020 building Lexant in the oncology space. I would say it's one of those times where liquid biopsies were the rage in diagnostics, especially if you would remember in the oncology space, that was all the rage at that point.

and specifically in tumor profiling. And then of course the market started to shift towards monitoring. we certainly saw a similar type of a, I would say investor and risk environment, which was quite different towards the end of 2020 versus the early times when we first got founded, like in 2016. So I think one of the things I found or I find fascinating is not only are you

sort of projecting where science will take your product, but you're also sort of projecting where the market is going to go in terms of its appetite to invest in these assets. And I think given a 10 to 12 year timeframe, which we've agreed that is the sort of the time you have to kind of work on to go hit the target. I think it's pretty enormous challenge, I think, from a biotech.

founders perspective, at least that's one of the learnings I've had from being a tech entrepreneur and then moving into biotech, right? Would you agree? Like, I feel like there's another order of magnitude of complexity involved in getting this right.

Alex Mok (18:23.574)

Yeah, yeah. I think it's building a drug and getting it through the trials, clinical trials and getting it approved is one of the most difficult things one could aspire to do. The level of, you know.

Ram (18:35.278)

Yeah, but at the same time, really impactful, right? I mean, I think it's probably one of the most impactful things one could also do, I feel. Yeah.

Alex Mok (18:40.856)

Yeah. Yeah. And it is, I, you know, I've had a lot of, uh, personal, uh, kind of, I don't want to call them perks, but just the feeling of working on something that can make such a big impact and reverse disease and prolong life and improve quality of life. Like those are the things that got me up every day. Got my whole team up every day, worked long hours, uh, scratching our heads on puzzling science, scratching our heads on puzzling investor conversations. mean,

The reason is because we want to deliver these medicines to patients who just can't afford to wait anymore. a single molecule, like a single structure, can affect billions of people's lives. And that is a really, really awesome, amazing thing.

Ram (19:24.994)

Yeah. And I think one of the really interesting comments you made is, 2017 to 21 is a very different market to 25 and beyond, right? And, you know, if you sort of look around and read the news and there's so much excitement around, especially AI driven tooling, AI driven techniques, models, which can do better drug discovery or optimizations or

So what sort of, know, given that you've done this pre-AI, I would say, or at least the maturity of AI, what's your view right now? I mean, know, pre-actual AI, pre-AIL labs, that's right. Yeah. Or rather like, you know, I think the one way I look at it is there's like what, like,

Alex Mok (20:06.68)

Pre-actual AI, but we still call it AI. A pre-Elo lens.

Ram (20:23.96)

trillions of dollars going into building AI and AI infrastructure. There's sort of like a wave which is rising all boards, sort of a situation going on. But I am curious, know, like given that you've had this perspective from the pre era, what does it look like right now? What are you excited about in terms of like what's happening right now? What are the type of companies which you think are going to be the next big leaders in the space, right?

Alex Mok (20:48.63)

Yeah. Yeah. I mean, I think there is, as you know, an infusion and distribution of AI all over intelligence is entering every nook and cranny. We're still figuring out, I think, the first iteration of applications. And I would say that from a drug discovery perspective, I mean, we all hear about Alpha Fold 2, Alpha Fold 3, all these other models coming out. mean, you know, the pace of development on the...

biological foundation model, the biology foundational models is also incredibly fast, right? It may be not as fast as, you know, open AI and Gemini and which it seems to be every two weeks or something, there's some new developments, transformational development, but you know, it's at a clip that's, you know, quarterly. And I think that that is bringing just a democratization, you know, deprecation of costs.

Ram (21:27.544)

Yeah.

Alex Mok (21:43.212)

for running certain studies and, you know, lot of science is, you know, how do you build conviction, right? Like anything else, how do you build conviction? And if you can build conviction through in silico modeling that you can begin to trust more and more and more and more, yeah, I mean, things are gonna really speed up, but that doesn't, you know, solve the clinical trial aspect, which is where the valuations really step up. I mean, we think about biotech in terms of stair-step valuations, it's all...

really based on meaningful clinical catalyst. If you look at evaluation, I'll just take an example. If you look at companies who report a phase 1A safety, that valuation between the phase 1A and a phase 2 can be literally 20X. I think it's, obviously you need to demonstrate you have mechanism, proof of concept in the clinic that you're doing it in actual patients, but just getting to the clinic.

which may take a long time, especially in new modality that I was familiar with, like exoskeletonal therapy, that's really just the third inning, I think, of the game. And we definitely need the technology, acceleration, just what Tilda's doing in the other six innings, right, to get to approval. And one of the things that I'll say is that it's not just...

running your clean ops efficiently that makes or breaks the company. Obviously that's very important. You clean dev strategy, clean ops strategy tied together, very important. But you have to think about how you're lining up your catalyst, what your competition is doing. You're going to need to predict based on, you know, asymmetrical information, whether or not your competition is going to read out at a specific time before you. And that could actually make your drug program completely move, no matter how much capital you put into it. You need to think about your financing runway. You need to think about

Now I've learned that you really need to think about macros. You really need to think about what the Fed is thinking. You need to think about where pharma is investing in their themes. Do they have drugs like in today's world where many loss of exclusivity, a lot of drugs are going off patents? What are they looking for? What are they shopping for? At what stage are they looking to buy something? Are they looking for a phase one asset today? Probably not. I could tell you most certainly not.

Alex Mok (24:05.618)

Are they looking for a phase two, phase three acid? Well, it depends. It depends on the indication. It depends on what your drug is. So I think, you know, the execution side, the strategy side in biotech is so important. And I just want to make the point here that, you know, you want to behave. One of the things that I think is most difficult about clinical development is the lack of control of the timelines for these very, very meaningful catalysts. mean, every week makes a difference.

If your enrollment takes, today, I think the stat is over 90 % of programs fail to hit their enrollment windows. that's everybody. And if you can't hit your enrollment windows, you can't start your trial. If you can't start your trial, you can't get the data, then what? You don't get funded. And then your company's gone. So I think that, yeah.

Ram (24:43.778)

Yeah, everybody, literally.

Ram (24:53.154)

Yeah.

Ram (24:58.54)

But right, you know, if I think about it from a, you know, first principles, right, assuming we're all trying our best, the industry is really like trying its best, investing in the right places. I do think there's clear challenges which we probably cannot immediately overcome, right? Like, you know, let's say enrollment windows being hit, maybe you're aggressive in your timelines, maybe there's not enough patients out there who can participate in all the trials which are being run on the same indication. I think there's some like other factors which probably also contribute to it.

But I certainly think that the bottlenecks of today may not be the bottlenecks of tomorrow or potentially we're like moving away from the way we think about these bottlenecks, right? Like is that kind of how you're thinking about it with the advent of new technology, new AI systems, automation, you know, we call them AI teammates, but in general, I think you're sort of seeing a shift in your bottleneck analysis or where you think those bottlenecks are, right?

Alex Mok (25:46.146)

Yeah.

Alex Mok (25:53.88)

Right, right. you know, obviously for any technology deployment, they first go to the bottlenecks, right? And so as those affect the bottlenecks and modulate them, we have new bottlenecks or, you know, things that get down the line, you know, in other arenas. So yeah, I would agree with that a hundred percent.

Ram (26:14.958)

So recently, I'm not sure if you're following this discussion thread, but lots of chatter about how there's an output gap. There's all this progress around AI model intelligence. Like you said, there's a new model which is trying to compete with the state of the art every two weeks. And these big model front-to-lap companies are competing against each other. You could call it a massive arms race, right?

But we're clearly not seeing the, there is an output gap and there is a lack of diffusion of all that great intelligence, I would say into a tangible form where you and I can actually see it meaningfully in our daily lives or in our daily careers and the things we're trying to do, right? So I think my perspective, or at least the question I'm trying to get to is, how do you think about that? And do you feel like, you know,

it's just around the corner. Are there reasons why that's not happening? Are those reasons going away? Do you have a perspective on where is this great amount of technology and progress, at least on the model intelligence side? How is that coming into the space where we can actually see meaningful impact on some of these bottlenecks we're talking about? Because ultimately, that's all that matters, right? From a value creation perspective in biotech.

Alex Mok (27:33.708)

Yeah.

Alex Mok (27:39.914)

Yeah. Yeah. I mean, I think for me personally, and some of my colleagues that I work with day to day, mean, we're utilizing tons of AI tools. mean, like most people now, think the number of the first, the first terminal I pull up is either GPT, Gemini or world cloud or whatever to answer any question I have, literally any personal or, or, or business. Right. And, and, you know, the models are adding value. The progression of the matters are adding value in all of those domains.

Um, um, but it's, it's, it's, it's becoming, you know, uh, you know, a thought partner that is generating outputs that I can then utilize. Um, you know, and it's, it is, it is, you know, I think, I think, uh, one example is just where slide deck making has gotten. It's kind of, it's just insane. Uh, in like, in like a few months, it's just really crazy. Like I would never use it for that. And now it just makes the decks for me and it's, I only have to make a couple of adjustments. Um, you know, so.

Ram (28:37.326)

What's your favorite platform for that?

Alex Mok (28:40.6)

I'm still exploratory. can't nominate one yet. I don't feel like I've done enough research. They all do it now, but Google seems to be doing quite well with some of the newest advancements.

Ram (28:45.516)

Okay.

Ram (28:55.672)

Yeah, I've stuck to Google, but I've tried Gamma and a couple others. think that's also an interesting close race. lot of interesting new companies in that space,

Alex Mok (28:59.797)

Yeah.

Alex Mok (29:04.47)

Yeah. Yeah. But, there are unlocks that I, I, I hear an update on LinkedIn or Twitter and then I go implement it and I said, wow, it did, it does that. It's literally every, know, if it doesn't do it today, then in a week I come back and sometimes it does it. Right. so, so I, just to kind of get back to the question, I think that there is, there are definitely, value that is.

being driven by the current forms of AI tools and technologies, but are they the biggest value unlocks in therapeutics and bioform and biotech? I think, no, they're not yet. There is, but we all see it coming, right? We all see it. There's a path there and it's a matter of the right interface, the kind of workflow and to what level is it?

You know, how do you actually, it's sort of like an automation curation question there. It's not just necessarily automation for automation sake, but it's sort of like, where does it make sense to actually build these learning tools? yeah, I mean, yeah, but I, I, I as someone who, you know, started a platform biotech company and ran into a major, you know, funding period that was

very challenging to kind of fund multiple programs to the clinic and kind of had to settle on one. If we can reduce the cost of clinical development from today, would have, I can tell you, 50 to 60 million to get to a POC of one proof of concept for one drug. If that can be reduced just by 30%, then the financing risk

is relieved significantly, 30%. And you can think about other ways for other programs to come to light and it gives you just more options. But if it's 60 million to get to POC and that 60 million can miss its milestones because of things out of our control, just because we don't have the visibility, just because we're not getting the real time data from XYZ CRO.

Alex Mok (31:27.544)

That's a real existential problem that I think needs to be solved.

Ram (31:32.558)

Yeah, and actually there's two topics which I find really interesting to talk about, right? One is this scenario of what you just said, tools are helping, but not to the extent we would like or we would want to be yet, but that's coming. I think we should double click on what that means and where we think that can happen. And second is what you just said, which is how do we kind of go from, you know, this like $60 million pipeline giving us certain risk, risk rewards.

Alex Mok (31:48.216)

Mm-hmm.

Ram (32:02.162)

And if we were able to increase or optimize that risk reward, what does the future look like? Because I do think the market is going to look really different once that happens. Both are sort of intertwined, right? Like the diffusion of all this power into some of these like bottlenecks will create a better ROI model. And I think that ultimately leads to a really different type of a biotech market is sort of how I see it. And I think my pattern matching here is

what cloud computing did to startups. You know, especially the risk reward involved in creating a startup in the tech space, which are, you know, like I'm talking about like a software as a service style company, right? Like a typical company of that sort. And the amount of risk involved in basically just trying something. And what does it take to build a, you know, a meaningfully large business and

Alex Mok (32:33.528)

Mm-hmm.

Alex Mok (32:44.664)

Mm-hmm.

Ram (32:58.53)

How does the ecosystem react when your risk reward sort of suddenly changes? And what does that mean to the whole, the entire economy around that, right? So I'd love to double click on all this. I'm sorry about that, but kind of going back to that diffusion part of it, right? One of the things I personally believe is,

You know, like the tools and chat, GPT and these kinds of ways of interacting with models, amazing. They give me a lot of personal productivity as you're also explaining your own workflow. But I think for an enterprise to really gain that sort of power in a daily workflow, you know, in like the day-to-day things we all do, we collaborate on, kind of work towards a milestone. I think that really needs to be

delivered in a different form and a different package. I don't necessarily see that as a chat interface. I think that's more of a fully packaged AI model with various sorts of interfaces and knobs and UI. It's got to be something different, something which is easy to adopt and easy to integrate into your current working model.

Alex Mok (34:17.079)

Mm-hmm.

Ram (34:18.018)

You cannot integrate charge-upd into your CTMSs or your EDCs or your forecasting models or your trackers. I feel like that's where the enterprise diffusion is like hard to do because enterprises have a certain hurdle rate before they can adopt technology. need to fit into, I mean, the basics, like you need to fit into that identity model. You want to log in using your company's accounts, username and password. And, you know, that's not possible with many of these AI tools right now.

And maybe once you bring them in, you should be able to tell them, use these five technologies I'm using already, and you help me go do this work. It's kind of like another person coming in and you're pointing them at a couple of logins and such, and then they can just pick it up and then go from there. So one of the perspectives I have is this sort of a vision naturally lends itself to the agentic future. We're all probably getting into, you know,

pretty much an accelerated path at this point. And I am excited to see where and how that impact will be, particularly because they solve for that, I think the output gap, that at least should be better than what it is today. So I'm interested in your thoughts, right? Particularly since we're talking about clinical trials, like, know, do you agree, do you have thoughts on like where that can happen? Where do you think the first set of impact will be?

Alex Mok (35:42.454)

Yeah, totally. I mean, maybe I'll give a pedantic example. you're like, worked with an IT of a company and someone, you know, you log in and then they take over your computer and fix your IT problem. I mean, I think, I think that's what we do. Yeah. Yeah. Yeah. You let someone take over your computer, whatever technology it was in the last few decades, VPN, whatever. but, then it just happened. Right. And you're like, okay, it fixed it.

Ram (35:53.632)

Yeah, I think I played that role multiple times with every family gathering, Alex. You might be talking about me.

Alex Mok (36:12.386)

Today we're relying basically, we're going through our workflow, we hit a block and then we like go to an AI with separate terminal, we ask it for input. It gives us a suggestion. then implemented our, you and so this, this kind of copy and paste fatigue, right? Is, sort of what we're all doing. But I think to your point is having this in a fully native platform, AI native where we're, we're in a workflow along with AI and you know, to what extent that looks like. I think there's many other creative.

visuals of what that might look like, but in essence we want there to be no boundaries and we want there to be you know Data memory from one set of Terminals to the next set of terminals to the and that's where I think the emergent properties come about that can really unlock, you know key issues because And and and again there's not

not divorced from the constant kind of pairing and partnering that a human and a human intelligence and an AI intelligence will happen on the same interface. But there needs to be, you know, a fully native interface. And I think, you know, it sounds like that's something that Tilda's building. That's something that I know others are building. You know, the browser has become kind of the battle space for AI because that is an active kind of sort of, you can imagine, an infinite dimensional sort of terminal where two

we can interact, right? And I think more of that will happen, but it will take some key and clever infrastructure development and design in order for that to happen.

Ram (37:49.272)

Yeah, you you use the word emergent behavior and I love that term because I think that's sort of one of the most fascinating thing about these AIs, right? What we're realizing is the way these AIs work today, they truly display emergent learning and emergent behavior. You you give them 10 examples and, or I mean, 10 million examples and somehow somewhere they're sort of figuring out a, you know,

some kind of low-level manifold where you don't need to explain all the rules of physics. It just sort of watches enough YouTube videos and is able to create another video out of it, which sort of mimics any other physics, like an object falling or something like that. So it's really fascinating. And that's been our learning as well. It's basically bringing these AI teammates, get them into a place where you can actually share, again, using your own words in a boundless way.

the information or different workflows that you're working through. In our case, it's clinical trials. And there is true emergent behavior which appears, which is fascinating. You see these things get really smart, really much better than whatever you could possibly sort of do. And I'm really excited because I think that type of behavior is only going to get better with better models, better contact sharing as we integrate into more more systems, as we do more more work with these systems.

I think there is a certain amount of collaborative power which we're going to get, of course, which will help us to be more productive. But I'm also considering this other vector, which is what happens when you have a really intelligent behavior which shows up and you're thinking about things you've never thought about or couldn't have thought about. And I think that currently I see things like, can accelerate some of your timelines because you're just

catching some of the mistakes earlier than which you probably would have. That's a quality improvement which shows up in a benefit from a timeline perspective in trial management. if you're like, for example, we manage regulatory or we managing site startup and site close out and site management, the entire site management functionality, DM functionality, data monitoring functionality. I think there's a lot of like interesting things which you sort of start to see in this kind of integrated platform.

Ram (40:14.842)

And I think that sort of leads to that, again, going back to the risk reward. think then you start to see better risk reward emerge as you build and adopt more of these technologies, right? I was thinking about another point you made earlier, which is, is biotech sort of like a winner take all or first takes all type of a market? Like not only are you competing with your competitors, is it one of those things where if you land

later than your to go sort of like your competitor in the trial process, is there a significant downside in some sense?

Alex Mok (40:55.028)

Surely. Yeah. I mean, when you think about market launches, getting your drug, if it's going to be better than standard of care, you know, within the same timeframe as another drug is, you know, let's say you're four years away. mean, treatment paradigm can really shift in that time and the adoption curve can really be flattened out of your new drug, even if it's better. Right. And I think

You there's obviously patent life, which we all have to manage in biotech. If you have 20 to 25 years of patent life, like every single year matters. Cause if you're, if you're not on market for that year, you know, if you have a real good blockbuster drug, if you have a mega blockbuster drug, could be billions, tens of billions of dollars that you're not there. So yes, it's, something that, you know, I think folks in our industry think about a lot. you, you want to be first.

You want to be best, know, this sort of first in class, best in class mantra is really what's repeated quite often. Many companies don't want to do anything that's not first or best. because the, when you, when you factor in, if you're second or you're me too, with all the other risks that one runs into could be science, could be regulatory, could be market, could be, you know, operational risk like that.

that our MPV, the risk adjusted MPV starts to diminish just very, very quickly. And I think timelines matter, like in every industry, but it's something that we have to monitor like a hawk.

Ram (42:31.234)

Yeah.

Ram (42:35.928)

Yeah, that's interesting. So basically your risk adjusted NPV is highly correlated to timeline, obviously, because of the cost and everything else. But it's also correlated in the sense that you may not be first in class or maybe you are best in class, but you're certainly not first in class if you don't manage that well. And so there's a pretty large impact, I would say.

Alex Mok (42:58.454)

Yeah, yeah. you know, in drugs, really focus on, our MPVs, course, peak your sales. And your peak in your sales need to be substantiated on market penetration of your drug. And if you're late, your market penetration is going to take a big hit. And you may not ever recover from that.

Ram (43:16.376)

Yeah, yeah.

Yeah. And, you know, just, just segueing into the clinical trial operations part of it, what's sort of your view on, you know, what, what the industry is missing or, the opportunity ahead, right? Just, just in everything which you've discussed, I know you think you're one of the deep thinkers I know from the AI front of things. So, so we'd love to get your perspective on what you think from a impact, right? Like where do you think the things

Alex Mok (43:34.498)

Mm-hmm.

Ram (43:48.622)

the markets going on the clinical trials operations side.

Alex Mok (43:52.428)

Well, I can kind of share a personal story. And I think you know this. So my father had stage three B cancer in 2020. We was actually diagnosed during COVID, which was a heck of a time to get diagnosed. But we went through a variety of different treatments. Nothing was working. And so the next option was clinical trials. And he had a rare stomach cancer. And that made it limited in the number of trials that were available.

We didn't have the option of going to phase three trials because there weren't really phase three trials. Phase two trials basically non-existent as well. So we're looking at phase one trials. And any patient that considers phase one trials as a drug developer is a saint because they're trying something that is unproven. They're putting their life on the line because they need another option and there's a belief in hopefully, you know, what the drug can do for them. And I scoured the earth.

You know, at that time I really wished LLMs, this was 2020, 2022, 2023. So the things I could do today with LLMs, you know, would have made this a lot easier, but basically created, you know, a dossier of 30 or so trials, manually emailed, called the clinical trial investigators, you know, the CRCs or what have you to try to get involved, try to get in. And, you know, on my side, I had done all my reading on the inclusion and exclusion criteria. I knew whether or not he was a candidate.

And it was just a matter of getting through the gatekeeper to get screened, to get started on a trial and find the timing it goes. And I'll just share one example of that, which was we were working with a major academic medical center, very, very renowned academic medical center. You know, I would have thought them to be, have the technical kind of technology supports to enroll. We were going through an enrollment process with one of the trials that they had from my father.

And the day before we were about to fly out to do the screening, was the point was the next day, they said, okay, yeah, so, and so then, you know, once we do the screening, then we'll give him the drug and they named the drug. And I said, that's not the drug. That's not the drug that we're talking about. We're talking about this drug. Here's the number of the clinical trial, NCT, blah, blah, blah, blah. And he said, that's not what I have you in here for. I was like, we just spent four weeks and you're telling me you couldn't even get this part right.

Ram (46:15.084)

my gosh.

Alex Mok (46:18.368)

And so we didn't go obviously to that center. found another trial that we tried to get on. And I think this is as someone who's, you know, started a company and our life and death basically predicated on these trials being enrolled by these are the types of things that are happening in the real world. And I think if you would ask anybody out there in clinical research in the U S or Europe, you would, they would not be surprised. Potentially. Maybe this is a, maybe a particularly egregious case, but I mean, people are writing things down on

notepads and they're trying to recall from memory and it's not, it's not, as you know, it's a very complex issue and it's not because they don't, they're not diligent, it's because you're overwhelmed, right? And they're overwhelmed with all the other pieces that they have to do. And as I, as I kind of, you know, had that very personal experience and then also tried to look into this space a bit more, I just realized the bar is, is, is, is on the ground right now. The bar, there is no bar that's above the ground right now.

And so we're raising the bar. Let's just raise it enough where we can't stomp on it anymore. And I think that that's kind of where we can go. And I think that that will have an incredible amount of impact, you know, that is unappreciated. and, I think, you know, when, when, when I, I, when I think about it from, again, a drug developer's perspective, if I can have any more confidence that enrollment window or the rates that I.

kind of predicating my timelines on are real and I get real time feedback on that. And I know, you know, from a clean ops perspective that all the side activations are going well. And I have a bit of an engineering mindset where my assumptions are not like a standard deviation of like infinity. My standard deviation is like, you know, a few days here and there I can live with that, right?

Ram (48:11.406)

Yeah. Yeah. As you said, the bar's basically on the floor. Just needs somebody to step on it, right? No, I... First of all, I'm sorry to hear that you had to go through that kind of process. I understand where you're coming from. And one of the learnings I've had just interviewing sites over the years is the problem is just not at the site, right? It kind of...

Alex Mok (48:16.216)

Yeah.

Ram (48:39.66)

goes all the way and broad and vertical basically because one of the things sites, one of the biggest complaints sites talk about during activation is I don't get the type of responses I would want from my sponsor or my CRA or whoever their Clunox team is. And the answer, mean, and then you go one level above and you ask them like, how come? no, why are we not able to approve a certain patient into a certain slot or?

Alex Mok (48:44.696)

100%, yeah.

Ram (49:08.524)

Why is the site not activated yet? They have four patients lined up. They're waiting to get into this thing. And it's just taking like months and months, right? To kind of get started. And you know, the reality is that they're overwhelmed too. They have too much to do. There's so much going on and things change and you know, it's a tough job. It's a tough, tough industry to work and really get stuff done this way. And that's one of the also like, feel

Alex Mok (49:16.471)

Mm-hmm.

Ram (49:38.03)

the benefits and of course the most exciting thing for me is what happens if you had some kind of a technology which is essentially being is immune to being overwhelmed. know, it's not the word overwhelm is not something which you would even understand, right? It's infinitely scalable. It's these properties which makes it infinitely consistent, you and you're talking about standard deviations.

I can talk about consistency rates and I can talk about alignment rates and measuring it continuously, monitoring it consistently. So I think there's a paradigm which I am really excited about where I think all of us in the very near term, I'm talking in the months to years ahead, will be given the type of tooling and the infrastructure which none of us had when we were starting our companies in 2016. Like when you did it in 17 and I was 16.

I think that's what's the most exciting part. I'm kind of coming back to that Ristavard equation, right?

I feel there's a lot of parallels and what's available to pure tech entrepreneurs with the Amazons and the Google Clouds of the world and so on and so forth, where you have some quite magical properties. have infinite scale, you have no capex of sorts. can convert everything to capex. You have massive distribution, have all these things which are basically like...

taken for granted today by that whole industry, but that didn't exist just 20 years ago that didn't exist. Right. And I know it because I was one of the early guys at VMware where we were building the early like hypervisor technology and things like that, which, which eventually is what's having all this impact, I would say. So what I want to talk about, and I'm quite fascinated to hear your vision for this. Assuming some infrastructure of that sort existed in the clinical operations space. And we're talking.

Alex Mok (51:15.212)

Right, right.

Ram (51:38.702)

bandwidth and bottlenecks being an issue in the clinical operations, particularly in clinical trial execution, right? What does that make the next generation of biotech companies? And it need not be the cohort starting today. It could be something which is starting two years from now, right? In 2027 or nine, whatever it might be. But what does it look like? mean, if you go back into, I mean, if you go forward into the future and decide, you know what, now is the time to go into exosomes or whatever the new modalities and I'm going to go do something new, right?

Alex Mok (52:05.762)

Mm-hmm.

Ram (52:08.254)

I'm just curious, do you feel like there's a view you have which you feel strongly about or excited about on wherever we're going?

Alex Mok (52:17.014)

Yeah, yeah. mean, I think at first glance, you know, if one assumes that intelligence diffuses into, you know, these interstitial processes, right. And I say interstitial because as you know, workflows are, there's one workflow that's staggered on top of another one that's just to another one. And there's always some sort of a human component that needs to push it along. And if you don't push it all the way to the end, just never, the whole thing is a zero. So it doesn't get to the binary out.

outcome that you're looking for. Right. And I think that what, what, intelligence offers us in the form of AI is this way to monitor that, report it, improve it autonomously. I think it can only be done through agents. I don't think that that's something, you know, you always run into this, but you throw software at the problem, more software, you know, it'll just, I think, create more, more gaps, or at least the same gaps will, will, will continue.

Ram (52:46.136)

Yes, yeah.

Alex Mok (53:15.122)

maybe take on a different form. But, you know, these agentic autonomy automation will, I believe, simply help stitch those, that fabric together and ensure that end to end, is achieving its goal. As we all know, these systems are goal oriented. It's very clear that they can achieve those in other domains. And I don't see why they can't do that here. Right. And I think these are all finite.

finite processes that can be switched together. So what does that give us? That gives us a clarity on kind of where we are in terms of tracking to a particular goal. And it gives you this ability to roll up that clarity, all those tracking into these finite interstitial processes into, you know, hopefully a real Gantt chart that you can actually follow and it's not just reactive.

or just a prognostication of something that was emulated from a previous trial, is always different. And it gives you that real time updates that is not required human intervention. And I think that that really changes the way that we manage capital, that we think about cap risk, that we think about resourcing different projects. We can then spin up, spin down projects, hopefully easier.

You know, there's, there's, as you know, there's a lot of, think inefficient study startup costs, because of assumptions that were baked in, you know, I don't know from when, sometimes the reference point is like five years ago. Sometimes it's, you know, from a different city, different state, different set of patient population. Like those assumptions aren't, aren't authentic. They're not, they're not real guiding principles. So I think, I think if we, if we can find a 1%, you know, improvement through all of these stacks of

of things, we can actually accumulate quite a large efficiency gain. And I don't know what that number is, but I think that we could look back after it happened to say that that was transformative. And I think that having this level of...

Alex Mok (55:36.152)

Continuity, know sort of communications in between all these these layers will actually help this especially in clinical trials, you know clinical trials is an incredibly complex and and and You know long process and I think just having more of this kind of If you recall kind of Internet of Things which is still going on but Internet of Things is putting sensors everywhere, right? You put sensors everywhere. They're all communicating together

There's a sort of hive mind that then gets emerges from that. And then there's better decision-making. There's a you know, resource management, all of the above. And I think that's sort of the vision that I would be seeing analogous to something that might be a fully, you know, AI native agentic workflow that gets, you know, parallelized across many tech stacks.

Ram (56:30.898)

There's one point you made, which was about, I think the way I think about it is you have to close the loop. There's all these like institutional processes and so on and so forth. But one of the things which I think is really key for us diffusing the technology into the right place is having an agent and have the capability and the design so that it can complete an entire process and close the loop.

What I mean by that is in a 10 step process, if you just optimize the third and fourth step, you're not necessarily like removing the process altogether. You you kind of have to like go and do and then like take the entire process out and make it something which an agent can quite easily do and perfectly do. But I think what's exciting is there's quite a bit of that you can already do in clinical trials because there's, it just naturally lends for itself at the type of intelligence we have today in these models.

Alex Mok (57:07.874)

Mm-hmm.

Ram (57:28.83)

And I think, I mean, I'm sure you're already following this, but 5.1 just came out. There's a whole new benchmark around workforce tasks now, right? yeah, no, no, 5.2 is at 70%. And I think 5.1 is in thirties in that particular benchmark. So there's a lot of focus on going after workforce tasks, knowledge-based tasks and things like that. And we're going to, we're going to open up a new frontier altogether, which will be close to, you know,

Alex Mok (57:34.412)

Mm-hmm.

Alex Mok (57:37.816)

This is for 5.2 or 5.1.

Ram (57:58.51)

Clothes are even better than where humans can potentially perform. And so you can kind of see, even the front-end model companies, labs are going towards that sort of improvement cycle right now. We've done a lot in chat, we've done a lot in video, image generation, our task planning, task reasoning, all this kind of stuff. But I think now we're certainly going one composite level higher, which will be some of these knowledge-based tasks or workplace tasks and things like that.

I think we'll all benefit from this, including Tilda. And I think that will lead to better diffusion because you just end up seeing a lot of tasks which are quite tied in can now be sort of stitched out and actually completed. Like the agents are quite smart enough to figure out how to do these things, even though they're kind of gnarly to like take apart and figure it out, right? The other thing I think you mentioned, which is about technology, a lot of tech gets added and new software gets put on and things like that. I have come to the view that

Alex Mok (58:44.204)

Mm-hmm.

Ram (58:57.39)

The whole diffusion model for AI is probably not another software layer. It's not another set of new EDCs and so on and so forth. I think it's just one system of intelligence, as I call it. It's basically something which sits on top of all your existing stack on your software. I think you have to do that because that way you don't need to replace anything which is already in the market. You add an intelligence on top of it. The way I rationalize that is...

If we were to bring in an additional human resource and they're trained on a particular workflow, they're not bringing their own software. They're basically bringing their capability through their intelligence and their mind and hands and eyes and all this sensory rate, and then they're able to perform the task. And I think that's where I also am quite bullish on the whole agentic sort of solutions because that's exactly what you'd have to do and you have to build the tooling to be able to do that. So that's not smart.

Alex Mok (59:48.364)

Thank

Ram (59:56.43)

That's just not the model itself, right? There's things outside of the model which you have to get right so that you can actually perform that work and then go do that. So pretty interesting. I have one last question on this. Assuming all of this happens, then we sort of end up in a world where maybe there is a set of AI teammates you can work with and you can sort of start to run a lot of experiments. Where I'm going with this is I recently saw a statistic that

You know like white coding platforms like lovable or you know those similar kind They're saying we're gonna soon surpass You know in the near future a billion new applications being built every year a billion, right? I mean, it's basically a prompt and then that develops an app at this point. I think Anyone who has kids in even like even middle school nowadays seem to be seeing that where they're they're just like white coding these apps and building

building these interesting software applications, mobile apps, things like that. Do you think that there is an equivalent of that on the biotech side? we essentially at the precipice where we always think of these 20,000 trials per year or active trials as a large number, but are we just about to get into a territory where we're just not even able to imagine the scale of the type of work and experiments we'll probably be doing in the near future?

Alex Mok (01:01:21.194)

Yeah, I mean, so long as input costs and input resources start to continue to move down, I think there's going to be tons of emergent applications and drugs and technologies that can be trialed. think that, you know, in our space, there's been a lot of talk about AI scientists, which is not just a LLM telling you what experiment to run. It is an AI scientist that's connected to an LLM.

Ram (01:01:43.64)

Mm-hmm. Yeah.

Alex Mok (01:01:49.932)

that has a recursive loop with automated hardware. So you have these full-on robotic machinery that's pipetting all the reagent cells, what have you. It's getting prompts and inputs in terms of protocol from the LLM, and you're asking it to do recursive science for you. And I think there's been a few papers recently. There's Cosmos is one that has talked about this, but this is where we're headed for sure. The level of automation is there.

The programming is there, the integration capabilities are all there, the programming and such required. And I think that that's really possible where I think I remember in 2016, 2017, 2018, there were these folks that wanted to create the Amazon cloud for bio, Yeah, Emerald, Transcriptic, you know, and those were

Ram (01:02:42.584)

Yeah, Cloud Labs. Like Emerald or I think there's a few others already. Yeah.

Alex Mok (01:02:49.986)

you know, excellent ideas at the time. And I think the technology stack just wasn't there today that they're there. And I think that there, we will see companies emerging doing that already. And, that, that, that, you know, reduces the input costs, gets to a point where we can do tons more iteration and you don't have to be in China and, and, and, and, and, and, and you can get to, you know, hopefully distinct molecules, which you can then test, but then lives a question is, okay, once I have these molecules, we're talking about drugs.

Ram (01:02:55.918)

Yeah, yeah, yeah.

Ram (01:03:09.742)

Yeah.

Alex Mok (01:03:18.904)

Or how am I going to test these quickly in trials? Right. that's, I just, I always think that that to me is, is, is, is, is the issue that again, the bar is on the floor there. So, uh, we've been trying automation and not an automated scientist idea, you know, for, for decades. Uh, but like trial automation, I don't know.

Ram (01:03:42.754)

No, like you said, I don't think the bar is as high as an AI scientist. think these are, and case in point, the reason why we've seen the kind of interaction we are is because of that, right? You sort of prove some certain extent of them built from there. And you you brought up a really insightful point and you use the word China. This is a heartily debated, talked about topic right now, how the US competitiveness is quite not where it needs to be or is being eroded by.

by Chinese, let's just say cost structure or structure on their programming, program development. And at the same time, we're also seeing a of like a historical dose in terms of biotech funding. I maybe now it's improved, but we've kind of seen that over the last few years. And I've always sort of taken the view that, you know, that's not going to change just by adding more capital to it.

just because the markets kind of turn around and valuation sort of like turn around, all of a sudden all the work is back to the US is probably not the right approach. I feel what's going to get us out of it is actually like every other place we seem to be doing in the economy right now. I think some kind of a diffusion of AI into this kind of a process will be the way to get the type of ROIs we need so that capital comes back into the US markets and essentially like creates that next.

next big boom in the biotech space. I'm curious if you sort of see that as, what's your view? Is that feasible, plausible?

Alex Mok (01:05:25.812)

So sorry, I don't think I quite understand the question. Are you thinking in terms of how the US can retain kind of leadership and within biotech with all these innovations going on in China?

Ram (01:05:39.15)

That's right. Yeah. So what's the... Yeah, exactly.

Alex Mok (01:05:44.236)

I mean, so in terms of technology development, I think that will continue to be quite competitive. But remember, we have a very mature clinical research construct here in the US and in Europe and other Western countries. And the issue is that they've just gotten really calcified. So if we can...

If we can transform those, alchemy zones into something that is more efficient, can we be like CCP and drive people in buses to enroll in science? Maybe not. Maybe Elon can, but I don't know. What I'm trying to mean, just kind of tongue in cheek, is that if we can really shift the way a clinical research is done, I think there's...

As I said, it was so hard for me to enroll my father into this quite rare cancer trial. And he's probably the super small fraction of the population that they were trying to find, but they couldn't even just respond to us. And I don't think the problem is that there aren't patients out there, that there aren't people willing to enroll in these trials. It's the fact that the trial construct in itself doesn't serve the ultimate goal of the industries that we're trying to.

to improve. China has the benefit of not having all that industry kind of backlog or tech debt, right? They're actually getting to start from this new, relatively new, but newer than the United States for sure. there's maybe more flexibility in how things are done over there because they can implement in these new burgeoning cities that they have all these different.

Ram (01:07:26.68)

Mm-hmm.

Alex Mok (01:07:37.688)

new initiatives, whereas we're still trying to figure out how to cut the tape of the 1980s, right, in 2025.

Ram (01:07:46.573)

Yeah, Interesting. Well, Alex, that's been, I learned a lot. Thanks for being here, sharing and yeah, no, it's fascinating. I think the perspective you have is always really, really interesting to discuss. So thank you and hope you have a good holiday soon.

Alex Mok (01:08:13.42)

Thanks, Ram. This was really fun. Yeah, I always loved, I loved deming out about these concepts and yeah, keep fighting the good fight until I think there's just so much possibility and opportunity and I'd love to live in a world that, you know, represents what we talked about.

Ram (01:08:30.264)

Yeah, yeah. All right.