Latest / The BioHub - by Avetix / Episode 208 - Handol Kim - CEO Variational AI

Transcript

Episode 208 - Handol Kim - CEO Variational AI · The BioHub - by Avetix

• 0:06
  So welcome back to another episode of the Bio Hub and
• 0:09
  Avetics podcast. Today's guest is Handel Kim, Co
• 0:13
  founder and CEO of Variational AI.
• 0:16
  Variational AI is a Vancouver based company using generative
• 0:19
  AI to help biotech and pharma companies discover and optimize
• 0:23
  novel small molecule drugs. Since founding the company in
• 0:26
  2019, Handel and the team have built a platform focused on
• 0:30
  exploring chemical space and accelerating early stage drug
• 0:34
  discovery. Pleasure to have you Handel.
• 0:36
  How are you I'm. Doing fine.
• 0:38
  Thank you very much. Pleasure to be here, Max.
• 0:41
  Now really good to have you. I think we we'll just start off
• 0:44
  with variational for, for those who don't know you guys, just
• 0:49
  tell us a little bit about you and the company.
• 0:51
  Yeah, absolutely. So we're based in Vancouver,
• 0:54
  Canada. We are 19 FTE, we've been around
• 0:59
  for oh almost seven years. So we, we were founded in
• 1:03
  September of 2019 and we focus exclusively on small molecules
• 1:10
  and now expanding into sort of these adjacent areas such as
• 1:14
  proximity based therapeutics like glues and protax, but also
• 1:19
  novel payloads for ADCS. And what we've done is we have
• 1:22
  developed our own foundation model for a small molecule
• 1:27
  generation. We it's called Enki and it forms
• 1:31
  the basis of our platform that we use to help biotech companies
• 1:36
  from small emerging biotech companies that are looking to
• 1:39
  hit some kind of inflection point in value all the way up to
• 1:44
  big pharma. And we're pleased to announce a
• 1:48
  collab with Merck MSD in September of last year 2 program
• 1:53
  deal $349,000,000. And so that's great validation.
• 1:57
  And we heavily leverage AI, we're AI 1st and myself and the
• 2:03
  Co founders, we come from a company called D Wave Quantum,
• 2:06
  which is now publicly listed and is a quantum computing company.
• 2:10
  But we were working on probabilistic and generative
• 2:13
  machine learning. And so that's that's where the
• 2:16
  team came from. And we have purpose built our
• 2:18
  own foundation model for this purpose.
• 2:21
  It is not an LLM and it is not software.
• 2:25
  Basically what happens is a partner will define their target
• 2:30
  product profile, a preclinical target product profile and, and
• 2:33
  he will generate these novel structures extremely
• 2:36
  efficiently. And we have a very high success
• 2:39
  rate in terms of being able to generate leads like right out of
• 2:43
  the model. So you can skip hit ID, you
• 2:45
  don't need to do a screen. And then we go in and do lead
• 2:50
  optimization as well. And then we, we get out.
• 2:53
  We don't have our own pipeline, we don't have our own assets,
• 2:55
  but we're really there to, to, to help the biotech companies.
• 2:59
  The team is composed mainly of machine learning and chemists.
• 3:03
  We have computational chemists and medicinal chemists in order
• 3:06
  to be the ones that look at the output of the molecules made by
• 3:10
  our machine learning scientists and curate and collate and
• 3:16
  provide those to chemistry teams.
• 3:18
  So we are AI 1st and we love working in the domain of small
• 3:23
  molecules because it's very different than than biologics,
• 3:26
  where I think the adoption of generative AI has really taken
• 3:29
  off. Yeah, I think this, the small
• 3:31
  molecule pieces is interesting. What?
• 3:35
  What was it that originally convinced you that there was an
• 3:38
  opportunity for AI to make such a meaningful impact?
• 3:41
  Well, you know, we, we published some results and delivered a
• 3:47
  paper at Nurips back in 2019 where we showed state-of-the-art
• 3:51
  results in molecular property prediction optimization, where
• 3:55
  we outperformed labs from, you know, from MIT and Stanford and,
• 3:58
  and various other places. And we thought, hey, you know,
• 4:02
  let's go do drug discovery, not really knowing anything about
• 4:05
  drug discovery or biotech, just as naive sort of machine
• 4:09
  learning people and, and software people.
• 4:12
  And that's why it's taken us almost seven years to get to a
• 4:16
  stage where we are, it's, it's very difficult.
• 4:19
  So we went into small molecules because that was what we
• 4:23
  developed the, the algorithm and the model for.
• 4:26
  There's, you know, training data and you know, we, we actually
• 4:31
  wanted to 1st apply the approach to other types of molecules, you
• 4:35
  know, polymers or, you know, OLED materials, you know, metal
• 4:42
  organic frameworks, various other sort of smaller molecules.
• 4:45
  Because we knew that we didn't really know anything about,
• 4:48
  about biopharma, but then we realized that, hey, there's no
• 4:55
  data. And the only place where there
• 4:58
  is enough data to train on was in, in, in biopharma.
• 5:03
  And as a result, we said, OK, well, we're machine learning,
• 5:06
  we're not magic. So we need data on which to
• 5:09
  train. So we decided to go into, into
• 5:11
  drug discovery. And so that was really sort of
• 5:14
  the impetus you go where the data is.
• 5:16
  And, and at that time, the, there were no large language
• 5:19
  models that were really as performing and nearly as
• 5:23
  performing as there, there are today.
• 5:26
  And so you didn't see the sort of the take off of the adoption
• 5:30
  of sort of protein language models.
• 5:32
  And there was no alpha fold at that time.
• 5:34
  But also our approach was specific to small molecules and
• 5:39
  applying AI to biologics is a completely different
• 5:41
  proposition. And there's some great companies
• 5:43
  doing some incredible things in that that domain.
• 5:46
  But it's really a sort of a tale of two modalities, if you will.
• 5:50
  So I think that and now we see sort of crossover and we see
• 5:53
  some some really, really interesting, you know, players
• 5:57
  and models, you know, Co folding, etcetera, etcetera,
• 6:01
  which is really, really, you know, exciting and it's a great
• 6:05
  time to be there. But when it comes to hardcore
• 6:07
  generation of ligands that are predicted to and then actually
• 6:11
  in experimental validation, actually, you know, bind and
• 6:15
  have potency and, and, and selectivity and, you know, have
• 6:20
  the right PK and, you know, have a good talks profile, that still
• 6:25
  remains a fairly rarefied territory.
• 6:29
  I think a lot of people when they see AI for drug discovery,
• 6:32
  they just kind of clump everything in together.
• 6:34
  And if you kind of look at what what companies are doing, not a
• 6:37
  lot of companies are doing what we do.
• 6:39
  I mean, they're, you know, a small number of companies that I
• 6:42
  think are credible in doing what, what, what we do, which is
• 6:45
  the, you know, the generation of novel molecules.
• 6:48
  And then it's also compounded by, you know, all do these open
• 6:51
  source models, which are which are wonderful.
• 6:54
  But then you know, really limited by sort of the, the
• 6:57
  scientific, the machine learning science depth of your team,
• 7:02
  which is still relatively shallow in biopharma.
• 7:07
  Would you say that's been the main challenge in your growth
• 7:09
  and sort of where you're going as a as a company?
• 7:12
  Well, I mean, not for less for us than than I think, you know,
• 7:18
  pharma or biotech and, and largely it's because, you know,
• 7:23
  AI is still a bit of a, a black box, but it's changing.
• 7:26
  We're starting to see a greater sophistication and nuanced
• 7:31
  understanding of what, what machine learning is and, and
• 7:34
  where it works and where it doesn't.
• 7:36
  But by and large, you know, when we entered the scene, there were
• 7:39
  a lot of companies doing AI for drug discovery.
• 7:42
  But when you scratch the surface and look at their, their, their
• 7:45
  AI people, they were largely computational chemists or, you
• 7:49
  know, statistical, you know, bioinformations or bio, you
• 7:54
  know, so and which is great. And, and these are people that
• 7:57
  are expert in their domain and they apply AI as a technique
• 8:02
  which can yield some really good results.
• 8:05
  But at the end of the day, if you want true innovation in AI,
• 8:07
  specifically for the domain, you're going to only get it from
• 8:11
  the science part of AI. And I think that was something
• 8:15
  that a lot of people didn't quite understand that that
• 8:18
  machine learning itself is a science.
• 8:19
  It's an information science, just like chemistry and biology
• 8:23
  are sciences, no experimental sciences.
• 8:27
  And you know, the, the, I always say, like, if you have a really,
• 8:31
  you know, thorny chemistry problem, you're not going to go
• 8:33
  up to a, you're not going to ask a pharm pharmacologist or, or a
• 8:36
  biologist, you know, you'll ask a chemist.
• 8:40
  And so if the challenge is really making AI work for you,
• 8:44
  then why wouldn't you go to an AI scientist, right, instead of
• 8:48
  going to a chemist or, or, or, or a biomedical engineer or to a
• 8:53
  biologist? And, and I think that's sort of
• 8:56
  been the challenge because for the domain of biopharma to
• 8:58
  access that level of scientific ability has been challenging
• 9:03
  because as well as, you know, as, as, as the industry pays it,
• 9:08
  it, it pales in comparison to what the machine learning
• 9:11
  industry pays. But more than that, it's about,
• 9:15
  you know, the use of machine learning.
• 9:17
  And, and if, if you could take your pick of going to work for,
• 9:22
  you know, anthropic or, or, you know, meta or go work for, you
• 9:27
  know, a pharma where you're going to be stuck in an, in a
• 9:29
  corner of the, of the company and just running models and not
• 9:34
  really being able to implement them, then it's a pretty easy
• 9:37
  decision in in addition to the fact that you're getting paid
• 9:39
  like two or three times more, you know, working at a, at a big
• 9:44
  AI company. So I think there are a lot of
• 9:46
  sort of systemic issues with breaking in cutting edge AI
• 9:50
  research. And I think also that the a
• 9:53
  pharma is not necessarily optimized structurally or
• 9:58
  organizationally to do innovation in AII mean
• 10:02
  certainly, you know, you're, you're in licensing as a pharma
• 10:06
  many assets from biology, from biotechs.
• 10:09
  So you're already very, very adept and well practiced in, in,
• 10:15
  in outsourcing innovation and then acquiring it, right in
• 10:18
  order to do the commercialization, which is, you
• 10:20
  know, the where, where, where pharma shines.
• 10:22
  Obviously they still do core R&D and they have to, but at the
• 10:25
  same time, I think you know, at least half of all programs, you
• 10:30
  know, and or commercial programs in pharma are basically in
• 10:34
  license from biotech. So why on earth when they do the
• 10:36
  same for AI because AI is even is even second or third order
• 10:40
  from biology and chemistry. So if you're already doing it
• 10:43
  for biology and chemistry, which is core, it stands to reason
• 10:46
  that that you wouldn't do it for AI.
• 10:49
  And So what we found is that most pharma are doing AI largely
• 10:55
  to understand how to evaluate other partners, right?
• 11:00
  Because if you don't have internal capabilities or
• 11:03
  understanding, then then how do you evaluate A partnership?
• 11:07
  So, and I think in, in the first generation, you know, back in,
• 11:11
  excuse me, like 20/17/2018, you know, before COVID, I think
• 11:16
  there were a lot of deals were, that were done because no one
• 11:20
  really knew anything. And so, hey, it's the the new
• 11:22
  shiny thing. And I think, you know, in the
• 11:24
  intervening number of years, I think AI has made a tremendous
• 11:27
  impact. There are a lot of companies out
• 11:28
  there that are delivering, you know, tremendous value and
• 11:31
  innovation for, you know, pharma partners and for biotechs in
• 11:34
  general. And I think we're starting to
• 11:36
  see a subtle, but not so subtle change in understanding where AI
• 11:41
  works well and where it can, where it doesn't.
• 11:44
  And so we're seeing the, the adoption and sophistication of
• 11:48
  partners grow rapidly and and that's welcoming to us because
• 11:54
  we don't have our own internal pipeline.
• 11:55
  We, we partner with everyone. So we're we're picks and shovels
• 11:58
  for a is for drug discovery. And you raised some interesting
• 12:02
  points there as well. You know, a is you guys got in
• 12:05
  pretty early in terms of how the AI market shifts, but but now
• 12:10
  it's becoming increasingly crowded, right.
• 12:13
  Well, what do you think makes you guys different to to some of
• 12:18
  the other companies out there? Well, you know, there's, there's
• 12:22
  the actual science and you know, the power of the model and which
• 12:26
  is, you know, completely dependent on the expertise of
• 12:29
  the team. So, you know, we've got a a
• 12:31
  world class team, but then everyone says they do right.
• 12:33
  So that that's that's tough to differentiate.
• 12:37
  But you know, it is actually a lot easier to determine whether
• 12:40
  or not a machine learning team has the chops and the true sort
• 12:45
  of scientific skill, which is have they published a conference
• 12:49
  paper at a major, you know, machine learning conference like
• 12:53
  eye clear or nerves or CML, right?
• 12:57
  Did they come out of a reputable lab?
• 12:59
  Like is there, is there, is there grad work in machine
• 13:04
  learning or is it in a different area?
• 13:06
  Right? Because if it's not in computer
• 13:09
  science or, you know, applied bath or stats or, or
• 13:13
  computational neural systems or, or these programs where you're
• 13:17
  doing machine learning science, then you're not really a machine
• 13:21
  learning scientist, right? And so it's, and then you know
• 13:25
  what? Whose lab did you come out of,
• 13:26
  right? You know, did you come out of,
• 13:29
  you know, you come out of like, you know, the rich Sutton lab
• 13:34
  and University of Alberta, your reinforcement learning person,
• 13:36
  you come out of the banjo labs, you come out of the lacun like
• 13:39
  Quran, etcetera, etcetera. It's, it's very easy to tell if
• 13:43
  you actually all you need to do is is spend, you know, a few
• 13:46
  minutes to understand where people come from.
• 13:48
  Is it reputable lab? But then it, it's, it's, I don't
• 13:53
  necessarily see the, the, I think that's changing.
• 13:56
  But when people sort of otherwise, you know, we see a
• 13:58
  lot of people in biopharma, they go, Oh yeah, we've got a this,
• 14:02
  this AI expert. And usually that person will be
• 14:05
  quite cheap. They should go like, well, I'm
• 14:07
  not, I'm actually a computational chemist and I, I
• 14:11
  really enjoyed this AI. But it's, it's just kind of
• 14:13
  like, it reminds me of the Internet, you know, back in the
• 14:16
  late 90s where, you know, you have all these people who don't
• 14:19
  really understand technology and they go like, oh, hey, you young
• 14:24
  person, right, You're on this Internet thing all the time.
• 14:27
  So can you write me your Internet strategy?
• 14:30
  And, and this person's like, oh, OK, like I just, I just started
• 14:34
  playing with Mozilla. I just started doing this.
• 14:36
  I, I can do some Java, but you're our AI guru now.
• 14:39
  And then they go like here, you know, step forth our champion.
• 14:44
  And so we see, you know, some shades of that in, in AI and,
• 14:48
  and I'm sympathetic because if you're an experimental scientist
• 14:52
  and you're in drug discovery and you see all of this hype and
• 14:55
  your, your investors and your board are saying, what's your AI
• 14:58
  story with? And you're like, I'm, I'm trying
• 15:00
  to, trying to make a, I'm trying to hit rheumatoid arthritis with
• 15:03
  a small Mulligan. So it's a, it's a, it's a hugely
• 15:07
  difficult biology and chemistry problem.
• 15:09
  What do I need AI for it? It's like, OK, let me, let me
• 15:11
  get them off my back. And so a lot of sort of this AI
• 15:15
  stuff is performative, right? But I think we're moving from
• 15:19
  performative to, and we're also seeing when we interact with,
• 15:23
  with potential partners, it's less about like, why AI to like,
• 15:26
  how, like, how would we use this?
• 15:29
  How would we work with you or someone else?
• 15:32
  And so that, that is, you know, a pretty significant shift, but
• 15:36
  you know, it's, it's a, it's a really interesting time.
• 15:39
  I think we're starting to see a bit of an inflection point in
• 15:42
  terms of the adoption. So I guess getting back to your
• 15:44
  question though, what's, what's our differentiation?
• 15:47
  I mean, there's the science part and that's, that's pretty,
• 15:51
  that's pretty, you know, easy to understand.
• 15:55
  But honestly, it's, it's our business model, right?
• 15:58
  Had we gone in. So we, we, we were thinking
• 16:02
  about, you know, applying our model and key to develop our own
• 16:05
  assets and our own pipeline. We actually went far down this
• 16:08
  path, you know, a number of years ago.
• 16:11
  But I'm pretty convinced that had we actually, you know, gone
• 16:13
  through and raised the money and deployed the capital, that we'd
• 16:16
  be dead right now. Because, you know, AI and
• 16:21
  specifically generative AI in, in small molecule, you know,
• 16:25
  discovery and development. The discovery has been pretty
• 16:28
  well validated. I mean, there are many companies
• 16:30
  that are showing, you know, tremendous efficiencies and
• 16:34
  success in the early stage. And I think even, you know,
• 16:37
  people who would would actually be, you know, rightfully cynical
• 16:41
  about some of this, like, you know, your Pat Walters or or,
• 16:44
  you know, your Derrick Lowe's or whatnot.
• 16:45
  They all concede like, yeah, yeah, OK, yeah, they can
• 16:47
  generate all these molecules, but that's not the main thing,
• 16:50
  right? And that's just a small part of
• 16:53
  the difficulty and the risk in doing drug discovery and
• 16:56
  development. And I completely agree.
• 16:58
  And further, you know, what is the probability of success of an
• 17:03
  AI discovered molecule in clinical development?
• 17:06
  And I think, you know, there's some data that says, you know,
• 17:09
  oh, Phase 1, you know, you have a higher rate.
• 17:11
  Well, OK, that, that's pretty, that's pretty easy.
• 17:14
  Just, you know, just file your Ind and, and hopefully you're,
• 17:16
  you're in the clinic. But then the the phase two sort
• 17:20
  of Cliff like it's, there's no increased probability of success
• 17:24
  in Phase 2. And that's largely because you
• 17:27
  know, it's not a problem that AI necessarily solves.
• 17:30
  When you have an AI developed molecule, you know, it fits to
• 17:33
  the properties that you define. And if the properties are wrong
• 17:36
  or the target is wrong, then you don't have a drug.
• 17:39
  And so we still the jury is still out in terms of what the
• 17:43
  probability of success in phase two and late stage clinical
• 17:46
  trials are. And until that moves
• 17:48
  appreciably, I think you're still taking a lot of risk as an
• 17:52
  AI for drug discovery company if you have your own assets, right.
• 17:56
  And we've seen this with several companies, you know, quite a few
• 18:00
  companies that are quietly or not so quietly, basically, you
• 18:03
  know, pivoted tremendously away from that companies such as, you
• 18:07
  know, your benevolence or your verges or whatnot, you know, ex
• 18:11
  ancias, etcetera, etcetera. I mean, AI, but as, as Alex
• 18:14
  Jaber Ankov in silico says, is like, yeah, you can, you can do
• 18:17
  all this stuff, but once you're in the clinic, you're moving at
• 18:19
  the speed of traffic. Completely agree.
• 18:22
  And So what you're doing is you have more shots on goal and you
• 18:24
  got like, you know, hey, maybe you got 20 programs that you're
• 18:27
  that you want to put into the clinic, but how the hell are you
• 18:29
  going to fund them, right? And So what you're what you're
• 18:32
  going to see is and what you are seeing is a lot, a lot more
• 18:35
  assets available for licensing and partnering from AI based
• 18:39
  companies. Because you know, at the end of
• 18:41
  the day, you're still, even if you raise hundreds of millions
• 18:43
  of dollars, you can only take one or two of these programs
• 18:45
  into, you know, into, into clinical development and then
• 18:48
  beyond. And I think really what you're
• 18:50
  starting to see is maybe a surplus of assets of quality,
• 18:54
  you know, in addition to what you can go to China and get
• 18:56
  right. But I think that's really the
• 18:59
  impact is it should be a, a buyer's market of assets for
• 19:04
  biopharmers. And I think that that part of
• 19:08
  the development becomes almost commoditized or becomes a lot
• 19:11
  more efficient than it's been in the past.
• 19:14
  And then so really it's, it's a, our, our primary thesis is that,
• 19:18
  yes, we have, you know, in my opinion, the best, you know,
• 19:21
  generative model for small molecule, you know, drug
• 19:25
  discovery. But at the end of the day, it's,
• 19:27
  it's our model that differentiates us.
• 19:29
  Because if you're a biopharma, where are you going to get AI?
• 19:32
  You, you, it's, it's very hard to hire people of sufficient
• 19:36
  scientific, you know, experience and knowledge to do that.
• 19:40
  And also they're going to be sitting alone and then, you
• 19:43
  know, forget about your AWS costs and this and that.
• 19:47
  We're just, we're just there to provide that, you know, in
• 19:50
  partnership, you know, with, with biopharmas.
• 19:53
  So I think that that is a right market for us.
• 19:56
  And I think that is our primary differentiation is that we've
• 20:00
  already done it and we're we're happy to to share that with with
• 20:03
  with. You know, with our ecosystem.
• 20:06
  Yeah, I love that. I love that.
• 20:08
  And if if you're a biotech who's currently brilliant discovery
• 20:12
  program, where does AI then have the greatest impact on them?
• 20:17
  You know, on timelines, costs, probability of success?
• 20:21
  Yeah. So I mean, you know we have
• 20:23
  several offerings at different stages of growth, right, and
• 20:26
  stages of the of early stage discovery.
• 20:29
  So at the very beginning, you know, you you're looking at a
• 20:32
  program, you want to do some early discovery.
• 20:35
  You know, the biologists have spent a lot of time, you know,
• 20:37
  teasing out an MOA in a, in a, you know, an indication and
• 20:41
  you've got some insight, right? And now you need to get some
• 20:44
  chemical matter. So you go to acro, you do a
• 20:46
  screen, you do virtual screening, you do all this or
• 20:48
  that. And you're, you know, you're
• 20:50
  trying to find something that gives you signal that yeah, I
• 20:54
  can, I can, I can find a molecule and then you Med chem
• 20:56
  your way to that. And so at that stage, that's
• 20:59
  that's where we come in. So instead of using the same
• 21:01
  starting points that everyone does who's going after the same
• 21:04
  target and the same indication, there's always multiple parties
• 21:07
  going after the same, you know, targets at the same time, but
• 21:12
  everyone starts from the same sort of starting points.
• 21:15
  Well, why don't you start with novel starting points and gets
• 21:17
  you into FTO, but also gives you differentiated molecules in
• 21:22
  different areas of chemical space.
• 21:25
  When we, when we look at drug discovery, there's the, the
• 21:27
  exploration phase and then the exploitation phase.
• 21:31
  And so exploration is really about like trying to find, you
• 21:35
  know, really, really different and novel and differentiated,
• 21:39
  you know, hits and starting points.
• 21:41
  And it's really hard to do that. You can't just randomly sample
• 21:45
  chemical space. It's just too massive.
• 21:47
  You won't find anything. And if you start so therefore
• 21:51
  you start from known compounds, tool compounds or things that
• 21:54
  are out there and then you go, OK, let's start here and then
• 21:57
  let's try and you know, you know, Med chem our way into
• 21:59
  novelty and let's, you know, scaffold top or let's, you know,
• 22:03
  do just, you know, walk the nitrogen around, change this,
• 22:06
  you know, our group or whatever, whatever.
• 22:08
  But as, as everyone knows that it's, it's a painstaking and
• 22:11
  long process and it's expensive. That's your exploitation.
• 22:15
  But people shortchange the exploration phase because where
• 22:18
  do you get novelty, right. So I'll just start here and then
• 22:21
  I move right into exploitation. That's why that exploitation
• 22:25
  phase. And, you know, hit the lead and
• 22:26
  lead optimization takes so long and has, you know, not a great
• 22:30
  rate of success. And basically programs die when
• 22:33
  you run out of budget and you run out of time.
• 22:37
  And, you know, assuming infinite time and infinite money, yeah,
• 22:39
  you know, the, the, the success rates would, would be very high,
• 22:43
  but they're not. I mean, you're in a competitive
• 22:45
  situation. So really what we're saying is
• 22:47
  that the exploration of chemical space, finding different areas
• 22:53
  to start, that is the really the key.
• 22:56
  Because when you find these different areas and different
• 22:59
  sort of, you know, chemical, you know, structures that other
• 23:02
  people are not starting from, the exploitation phase is
• 23:07
  reduced. And if you look at costs and
• 23:09
  timelines of discovery, lead optimization is obviously the
• 23:13
  longest and the riskiest part of drug discovery in pre in the
• 23:17
  clinical side. But if we could reduce that risk
• 23:20
  and reduce the time and the cost, right, then we think we
• 23:23
  have a very transformative, you know, value proposition.
• 23:26
  That's that's what that's what we're about.
• 23:27
  I mean, our our sort of explicit tagline is to redefine the unit
• 23:31
  economics of drug discovery. And so that that's really what
• 23:33
  we're focused on. So that I would say that that is
• 23:35
  our primary differentiation. We're not trying to make drugs.
• 23:38
  We're trying to make molecules that can become drugs for our
• 23:41
  customers, and we can do it about 100 times more efficiently
• 23:45
  than status quo methods. Oh, it's so true.
• 23:47
  That so, so true. You know, like there's a bunch
• 23:51
  of companies that are trying to recreate what you have in house
• 23:53
  as well, right? Like it's, it's interesting for
• 23:56
  sure. Yeah, yeah.
• 23:58
  Why, why? Why do that?
• 23:59
  We can do that for you and then you can focus on the hard stuff,
• 24:02
  you know? Exactly.
• 24:04
  So, so true. I mean, you, you think about
• 24:07
  variational as a company, it's taken 7 years per FET.
• 24:09
  What you guys have like it's, it's the right way to go, right.
• 24:15
  But yeah, look, it's been awesome having you on.
• 24:17
  Ando. It brings us to a final
• 24:19
  question, which is a bit of a change of topic, but it's a
• 24:22
  question we ask every guest. Umm, if you could have dinner
• 24:26
  with any scientist or innovator, past or present, who, who would
• 24:30
  it be and why? I'd be Nikola Tesla for sure.
• 24:35
  Love that answer. Yeah, yeah, I mean, it would
• 24:37
  just like what? What awesome, amazing weirdness.
• 24:42
  Umm, would, you know, sitting down with him and Mark Plain,
• 24:45
  you know, like in in New York at the time and maybe JP Morgan.
• 24:49
  But yeah, Tesla would be would be fantastic.
• 24:53
  And then I'd also be able to see if he was actually an alien or
• 24:55
  not. But yeah, Tesla for sure.
• 24:59
  Love that answer. Look, it's been a pleasure
• 25:01
  having you on handle. Please everyone like and
• 25:04
  subscribe to the bio hub and and thank you so much.
• 25:07
  Thank you, Max.