.avif)
.avif)
Google and Microsoft have both shipped models that predict new materials. Years on, most of those predictions have never actually been made in a lab. Dr. Andrey Ivankin, co-founder and CTO of Mattiq, calls that a reality problem — and he is building the hardware to fix it.
Watch the full episode right here. If the player doesn't load, watch it on YouTube.
Listen instead: Spotify · Apple Podcasts · YouTube
The short version:
- Materials AI is trained mostly on computation, and most of what it predicts is never made. That gap is the whole problem.
- Synthesis used to be the slow step. Put a hundred thousand materials on one chip and it stops being the slow step — characterization becomes the constraint instead.
- The previous generation of materials discovery companies failed by narrowing. Mattiq's bet is that going broad is what makes the model generalize, and the chip is what makes going broad affordable.
Episode one was the view from inside the federal system — Dr. James Warren at NIST, fifteen years into the Materials Genome Initiative. For episode two I wanted the view from the bench: someone actually generating the data everyone says the field is missing.
My guest is Dr. Andrey Ivankin, co-founder and CTO of Mattiq. He's a chemical physicist and engineer by training — "a rare and proud Siberian," as he puts it. In 2015 he co-founded TERA-print with Professor Chad Mirkin, building nanofabrication tools now used by researchers around the globe. Mattiq came out of the realization that those tools, combined with new chemistry from Mirkin's lab, could change how materials get discovered in the first place. The company has raised close to $25M in venture capital plus a few million in government funding, led by Material Impact, and is currently raising a Series A.
This one is personal for me. I spent six years of my PhD using polymer pen lithography, the printing technology underneath all of this. I can testify the technology works.
Editor's note: Quotes are condensed for length and clarity. Questions and answers have been grouped by topic rather than in the order they occurred. The full transcript is at the bottom of this page.
Why can't we discover materials fast enough?
Because materials science never built the machinery to run experiments at scale, and biology did.
You want better batteries, you want more energy efficient AI chips, you want to explore deep ocean, or you want to go to space — all of those technologies require new materials, and we just can't discover them fast enough.
The reason for that is that unlike in biology, materials science never developed experimentation at massive scale. We still live in this pre-genomics-chips revolution in materials science. And without that scale, you can't explore design spaces faster. You don't have experimental data at scale that you can use to train AI that will be consequential, and you can't rapidly improve that AI to accelerate the discovery. So we are stuck in that slow, expensive discovery cycle.
The comparison to biology isn't decorative. Andrey spent time in biotech and DNA-related research, and he's describing a transition he watched happen once already:
First there was serial experimentation, and then people started to do a lot of automation in biology — 96-well plates, 384-well plates, adding robots to run experiments faster. But the unit of experiment was still one well at a time. It's still slow, and to scale your experimentation in that case you need more robots. The throughput is not easy to scale.
When genomic chips arrived, now you have a chip with, instead of 96 experiments, 10,000, 100,000, millions of experiments on a single chip. And how do you scale your throughput? You just reduce the density. So you can run much more, it becomes cheaper, faster. That's what we are trying to bring to materials science for the first time.
What is a megalibrary?
A single chip carrying hundreds of thousands of different materials, each one in a known location.
I like to think about it as genomic chips for inorganic materials. We brought together massively parallel printing — a technology called polymer pen lithography — and combined it with innovation in chemistry that allows us, on one chip, to synthesize hundreds of thousands of unique nanomaterials, all positionally encoded. Then we screen them for any properties of interest, from structure to function. Whether it's a good magnet, whether it's a good conductor, some other type of battery material — at a pace you couldn't do before.
For anyone who hasn't held one of these: it's roughly a square centimetre, patterned with hundreds of thousands of tiny pyramid-shaped tips. You engineer the chemistry and the composition, print once, and you have a vast relationship database — these conditions and this composition produce this result — laid out in a grid you can go back and read.
I asked him how fast, in practice. His answer separates the two halves of the problem, and that separation turns out to be the most important thing in the episode:
We should separate synthesizing materials and testing them. In terms of synthesis, we can print chips in an hour, maybe even less. So in one day with one instrument you can print ten of those chips.
Characterization really depends what properties you go after. Some techniques — optical, or some electrochemical — can run very fast. In a matter of a few hours you can characterize a hundred thousand new materials. Structural characterization, some additional functional characterization, those can be slower, and you need to be wise about how you characterize. But nevertheless you can run at a much faster pace — orders of magnitude — than with any alternative approach.
How is this different from a self-driving lab?
It changes the shape of the loop, not just its speed.
Self-driving labs — the subject of most of episode one — close the loop between computation and experiment: an AI proposes the next experiment, machines run it, the result feeds back. Andrey is careful that Mattiq is inside that category rather than opposed to it. But the loop itself changes when synthesis stops being a step you wait on:
The traditional loop is: generate hypothesis, synthesize one material, test it, learn something from it, repeat. That loop can take, in a traditional automated workflow, hours — in some cases days per material. And which step is the longest? In most cases it's actually synthesis.
With a megalibrary, you synthesize a hundred thousand materials at the same time, positionally encoded, and your loop becomes: generate hypothesis, go measure, learn, refine hypothesis, repeat. So you eliminate the longest step from your loop almost completely — I mean, you still need to synthesize new megalibraries occasionally. But that gives you the ability to test not one hypothesis at a time, but hundreds or thousands of hypotheses at the same time.
There's a real trade being made here, and it's worth naming plainly. Delete synthesis from the loop and you don't arrive at a solved problem — you arrive at a different bottleneck.
The bottleneck moves to characterization
Making the material was always half the job. The other half is proving the thing you made is the thing you meant to make. Mattiq started with the techniques that read fastest:
Obviously you try to start with something easy and simple. In that regard we started with optical characterization and electrochemical characterization — a technique known as scanning droplet electrochemical cell. It's a small droplet where you run your chemical reaction, and you can raster that droplet across the chip and measure the activity of a given catalyst at every single point. We started there because those techniques are easier and faster to produce data. But that's not where we want to stop.
Where they want to go is harder and, he argues, more general. Atomic force microscopy is the near-term expansion, in partnership with instrument makers:
We work with Bruker and Oxford on expanding and developing the characterization and application of that technique to megalibraries. You will never have the throughput of optical microscopy, but you can run fast enough that it becomes very meaningful.
Atomic force microscopy I like to call the Swiss knife of characterization, because it can be operated in so many different modes. You can look at whether your materials are good magnets, whether they're good conductors, at mechanical or thermal properties, piezoelectric properties. All of that information becomes available.
And structure, which is the part that makes a material genuinely characterized rather than merely interesting:
You only have a material fully characterized when you have composition linked to synthesis, linked to structure, and linked to the properties that material produces.
Then the number that stopped me:
If you think about what's available today — humanity to date has characterized and documented in the Inorganic Crystal Structure Database around two hundred fifty thousand different inorganic materials. When you combine megalibrary and STEM in an automated fashion, you can surpass that in a matter of a week with one instrument.
Worth adding a footnote of my own here: the ICSD sits behind a paywall. Meta and Google have both open-sourced materials databases and described them as ground truth, but those are computational results. There have been cases where structures announced as discoveries were already documented in the literature — plausibly because the paywalled record wasn't available to the teams involved. The access problem and the reality problem are tangled together.
So what's the business model?
Not selling data. Selling the judgment the data produces.
Data by itself is not very useful. Data can provide insights, and the more data you have, the more generalizable your insights become across materials and across properties.
That word — generalizable — is where his read on the industry's history sits:
Our thesis is that most materials discovery companies failed because they couldn't generalize. What happened is they had to focus on a very specific vertical and a very specific application, and that doesn't scale very well. Our premise is that by going at scale and broad across applications and properties, eventually we have a model that can get us answers across applications fast.
And the beauty here is that we don't only have that model, we have the engine, the real-world engine behind it, that allows us to quickly validate those predictions and refine the model. So we arrive at the answer in very few loops.
In the near term that means a few verticals — catalysis and energy more broadly, magnetics, semiconductor materials — and partnerships rather than product sales:
We partner with companies who have materials problems, because our data has two benefits. It helps you find new materials faster, and it also provides guidance on which materials are easier to scale and bring to market.
We engage, share our capabilities, make sure we operate in the same domain in terms of materials and properties of interest. Then we define the problem statement — what material you're trying to identify, what you're trying to solve for. We define the milestones, the budget, and the capacity of the factory required to explore the universe you're interested in, and see if we can discover completely new materials and generate IP that's not out there. Once we generate validated materials on chip, we pass along the formulations and you validate.
We don't think we replace your R&D necessarily. But if you can do hundreds of thousands of experiments a year, we just give you the best chance that you'll succeed with those experiments and focus on the best materials.
The second benefit is the underrated one. A model that tells you which candidate is easiest to manufacture is doing something a materials model usually doesn't: it's answering a question the person with the budget actually has.
Why did the newcomers raise hundreds of millions?
I've tracked materials informatics for a long time, and suddenly there are newcomers raising hundreds of millions — at least an order of magnitude more than companies like Mattiq. I asked him what those companies are doing differently. His answer is the argument at the centre of the episode:
One interesting shift in the past couple of years: if you look at the GNoME model that Google DeepMind came up with in 2023, and then Microsoft came up with some models — everyone was excited, but the reality is we are many years later and most of those predictions have never been validated. If you look at DeepMind, they're building their own wet lab to run materials experiments. The field is realizing that we need experimental ground truth to train models and actually make something useful out of them.
Not to say that the models are bad. But I like to say that materials AI today has a reality problem. The training data is mostly computational, most predictions never validated. We need to change that.
The field understands the problem, he says. The difficulty is what fixing it costs:
If you try to do it by automating experimentation the way we've been doing it to date, you require a lot of CapEx and a lot of development. And not only development — because the whole field is not ready. Instrumentation vendors, the instruments themselves, are not ready for running in the loop like a self-driving lab. So you need to put a lot of effort into optimizing and developing APIs and so on. But also, to get to a meaningful scale, you just need a lot of CapEx if you're doing it the traditional way.
Which is a fair description of why those rounds are the size they are. Building physical reality into the loop is a capital-intensive act, and someone has to do it.
He doesn't frame the well-funded labs as competitors:
Going back to the analogy with biology — we don't see megalibrary as an alternative approach to what Lila or Periodic Labs or Radical AI are building. We operate at different scales. We work at small, nano scale, and the advantage we get from that is we can run much faster and more cost-efficiently. At the same time, what they do is a little bit closer to the end application and the format of materials, how it's going to be synthesized. So we think about it as a funnel.
A funnel is a more useful mental model than a race. The upstream stage searches an enormous space cheaply; the downstream stage takes the survivors toward something manufacturable. Both are necessary, and they fail differently.
On leaving the lab
The last stretch of the conversation is for anyone at a national lab or a university thinking about starting a company — which, given who listens to this show, is a lot of you.
When you become a founder, you think you'll keep developing — especially coming from a technical background. You think you'll spend all your time developing the most exciting technology yourself, writing code or working on hardware in the lab. The truth is there are a lot of things you have to do and worry about. Fundraising, hiring, interacting with VCs and your supporters. All of those are very important and necessary, and I don't necessarily say that I dislike it. What I'm saying is that the scope of your responsibilities expands very quickly from you being a grad student or postdoc.
In a way there's some similarity with a professorship, because again you have to worry about bringing in students, getting funding, making your department happy. So there are more glamorous and less glamorous things you'll have to do. But at the end of the day it's very fast-paced, it's very rewarding intellectually, and if you're an idea guy or girl who always comes up with many ideas — that's the best thing to do.
His practical advice comes from having had an unusually soft landing himself:
I was lucky when I started, because my co-founder Chad Mirkin was a serial entrepreneur by then. He had a lot of knowledge and we had a lot of resources to get us started. But I see many people face a lot of challenges when they're all first-time entrepreneurs and don't have the network and support they need to start a company. That can be slow, unproductive and painful. In that sense, finding experienced advisors or co-founders early on is very helpful. You can save a lot of pain for yourself.
My own version of that, for what it's worth: ask yourself honestly whether hiring people and worrying about fundraising are things you'd be willing — or even happy — to do, rather than things you'd tolerate. If the answer is no on enough of them, that's not a reason to stay in the lab. It's a reason to go find a co-founder.
Three things I took away
- The reality problem is the whole thesis. Computational training data with unvalidated predictions is a self-reinforcing loop, and everyone serious is now spending real money to break it — DeepMind with a wet lab, Mattiq with a chip, the newcomers with nine-figure rounds.
- Deleting a bottleneck relocates it. Take synthesis out of the loop and characterization becomes the constraint. Mattiq's roadmap is really a characterization roadmap, which is a more honest way to read the company than the AI framing.
- Generalization is a strategy, not a property. The claim that earlier materials discovery companies failed by narrowing into one vertical is a specific, falsifiable bet. Going broad is expensive, and only affordable if your per-experiment cost is low enough — which is what the chip is for.
Mattiq is at mattiq.com. They're raising a Series A and open to conversations with partners who have materials problems — Andrey's LinkedIn is here. The full episode transcript is at the bottom of this page.
Episode one, with NIST's Dr. James Warren on fifteen years of the Materials Genome Initiative, is here.
If this was useful, follow along wherever you listen — and send it to the person you think will argue with us.
AI Goes Deep is a podcast from FedTech.
Full transcript
Lightly edited machine transcript. Filler words are removed and names corrected.
Read the full transcript
Xiao Zhong
Okay, so welcome to AI Goes Deep again. So this is the show where Deep Tech Venturing meets applied AI from the people that are building it. My name's Xiao. I'm the VP of Venture Building here at FedTech. So today we're gonna go to the bench from the people who are making the data sets, who's actually using AI to develop new materials. Please join me in welcoming Dr. Andrey Ivankin, co-founder and CTO of Mattiq. Welcome to the show, Andrey.
Andrey Ivankin
Hi Xiao. Great to be here. Thanks.
Xiao Zhong
Let's actually start with some introductions about yourself. I know you are a serial entrepreneur yourself and you have the deep tech background. Can you tell us more about your background and also more about Mattiq? What the company does, how it got started and how much money you raised? Lots of questions all your way. Let's get started.
Andrey Ivankin
Yeah, yeah, absolutely. So I'm a chemical physicist and engineer by training. Rare and proud Siberian. Actually through my career I've spent a lot of time as a deep tech entrepreneur and operator, building companies at the intersection of nanotech, biotech, scientific instrumentation, and now materials science and AI for science. So the first company I started back in 2015 together with Professor Chad Mirkin. And the company was called TERA-print. And the mission of the company was to develop a suite of nanofabrication tools that we call TERA-Fabs to make nanofabrication and prototyping cheaper, simpler, and more accessible tool. And those tools are now being used by researchers now around the globe, across applications, semiconductors, electronics, advanced electronics, biotech, and the increase in the AI. But later in 2021, we realized that hey, some of the technology when combined with some of the new chemistry developed in Chad's lab, can actually fundamentally change how we discover new materials. And that's when we started Mattiq, and that's been my main focus. Mattiq basically solves the very simple premise that you want better batteries, you want more energy efficient AI chips, you want to explore deep ocean, or you want to go to space, all of those technologies require new materials and we just can't discover them fast enough. And the reason for that is that unlike in biology, material science never developed experimentation at massive scale. We still live in this pre-genomics chips revolution in material science. And without that scale, you can't actually explore design spaces faster. You don't have data, experimental data that at scale that you can use to train AI that will be consequential, and you can't rapidly improve that AI to accelerate the discovery. So we are stuck in that slow expensive discovery cycle.
Xiao Zhong
Yeah. And then I have a ton of questions that we can dive deeper and unpack. But first of all, I remember you raised a few rounds. So how much have you raised then from which investors?
Andrey Ivankin
Yeah, so we've raised close to twenty five million in venture capital plus a few million in government funding. To date in terms of investors. Lead investor is Material Impact fantastic investor and then CS Ventures is another big supporter of our mission to date.
Xiao Zhong
Yeah, yeah, that's amazing. You touched on quite a few points there. First of all, as you know, maybe the audience don't they don't really know here is my background here is also in materials. And then I personally use the technology you mentioned that was developed there at TERA-print before. So polymer pen lithography, that's a high throughput printing technology that can enable you to generate lots of printing. With different features, with different things, which is a perfect vehicle. Maybe you'll get into this when we talk about how do we generate data here at Mattiq. For that's to me, that's a few generations down when it comes to how do you use this technology to be as a sort of generating the so-called megalibrary. I'm sure we're gonna be talking about that as well. So as a person who has been using that during my PhD for six years, I can testify technology works. That's that little thing there. But then going back to the data, to the data aspect of things there, I totally agree. So we'll talk about the other companies doing similar things, or maybe similar but different things in when it comes to AI, self driving labs, materials discoveries, et cetera. But first let's dive deeper into Mattiq. So when it comes to an AI deep tech company, there's the data aspect, there's the computing aspect, there's the algorithms there. And it's especially when it comes to hardcore science, when it comes to data, there's computational data, there's experimental data. We're seeing a lot of open source the computation data points being available there. So what's your secret sauce? How are you, differentiating yourself there? How do you generate the megalibrary and then what that is?
Andrey Ivankin
Yeah, yeah, absolutely. So I like to think about it as genomic chips for inorganic materials. Basically we brought together the technology that you mentioned, massively parallel printing of technology called polymer pen lithography, and combined with some innovation in chemistry that allows us on one chip to synthesize hundreds of thousands of unique nanomaterials all positionally encoded, and then screen them for any properties of interest from structure to function, whether it's a good magnet, whether it's a good conductor, or some other type battery material, at a pace that you couldn't do before. And that's again very similar transformation. If you look I spent some time in biotech and DNA-related research. So I see a lot of parallels in terms of the development. First there was serial experimentation and then people started to do a lot of automation in biology, where they had 96-well plates, 384-well plates, and then adding robots to run experiments faster. But the unit of experiment was still one well at a time. So it's still slow and to scale your experimentation in that case, you need more robots. The throughput is not easy to scale. When genomic chips arrived, now your unit of now you have a chip with instead of 96 experiments, 10,000, 100,000, millions of experiments on the single chip. And how you scale your throughput, you just reduce the density. So you can run much more, it becomes cheaper, faster, and that's what we are trying to bring to material science for the first time. Similar chips. But now with inorganic materials that can produce the data at scale, experimental ground truth data.
Xiao Zhong
Yeah, again, I probably know this space too well. So you have one centimeter squared chip on top of that you can and literally engineer hundreds of thousands of little pyramid shaped tips, and then all you have to do here is to engineer the chemistry and the composition. So you can it's you're creating this vast relationship database where you have certain conditions, certain compositions there to give you certain results and then you do one printing. Usually I remember when I was doing the simpler version of alkanethiol as the molecule and on gold you can do it the whole thing in an hour or maybe two hours right and in theory how much time are we talking about to generate what number of data points there I have some maybe a few hours you can generate hundreds of thousands of data points. Is that kind of the speed we're talking about?
Andrey Ivankin
So yeah, so we should separate the synthesizing materials and testing them. So in terms of synthesis, we can synthesize chips print them in an hour or maybe even less. So in one day with one instrument you can print ten of those chips. But characterization it really depends what properties you go after and some characterization techniques optical or maybe some electrochemical, they can run very fast. And then in a matter of few hours you can characterize hundred thousand new materials very quickly. Structural characterization, maybe some additional functional characterization, those can be slower, and you need to be wise about how you characterize it. But nevertheless you can run at much faster pace orders of magnitude than with any alternative approach.
Xiao Zhong
Yeah. I know we're probably going off the topic here. I know we have a list of questions, but I'm obviously jumping, but this is so interesting. People in the space there's audience here, people with general backgrounds here as well. There's the term of self-driving labs, which is in the materials space, it's defined loosely as high-throughput experimentation. And then I know a bunch of people doing that, of course, as well. You do as well. So one of the pain points here is, well, first it's faster than a graduate student doing it, but then it's still much slower than what DFT computation or computational chemistry can do. But then of course, the trade-off here is for computation, it let's say density functional theory, which is the standard. It still lacks the it's still not the truth. So the ground truth, so to speak. But then when it comes to self driving labs, every almost every case it's it needs a bespoke build. And then still the speed is not as fast as what we want. And then on the other side we have the megalibrary or your solution from Mattiq. How do you compare these two hardware data generating approaches there?
Andrey Ivankin
Yeah, and two points there. First of all there is this concept of self-driven lab that you mentioned, and that applies both to Mattiq and every other company in the field that puts AI in the loop and helps generate hypothesis and synthesize materials that are the most likely to be interested in and test them so that happened in the loop. Mattiq and megalibrary actually changes that loop slightly because the traditional loop is generate hypothesis, synthesize one material, test it, learn something from it, repeat. And that loop can take in traditional automated workflow hours, in some cases days per material. And which step is the longest? Well in most cases it's actually synthesis. So with megalibrary, you can synthesize hundred thousand of materials at the same time, all positionally encoded, and your loop becomes generate hypothesis, go measure, learn, refine hypothesis, repeat. So you eliminate the longest step from your loop almost completely. You still need to synthesize occasionally new megalibraries. But that gives you significant acceleration and that gives you the ability to test not one hypothesis at a time, but hundreds or thousands of hypotheses at the same time.
Xiao Zhong
Yeah, no, that's very interesting. So and then for in layman's terms there. So when you there's the synthesis, which is you mix A and B, gives you C. So sounds like one of the differentiators that the your technology can do here is you can do this in parallel super, super quickly. It used to be a bottleneck myself, I was making molecules, it was really miserable. You do HPLC all the time. So it just takes a long time and then you have the yield issue to worry about. And then when you've solved the yield issue, you have to think about how do you scale up from this one, milligram to 10 milligrams, etc. But you solve that problem. You can do hundreds of thousands of those little almost little chambers of synthesizer doing this all together. So now you solve the A plus B equals C problem in a high throughput way. There's the other one, which is the you call that we call that the characterization, which is you want to make sure C you make is actually C. There are technologies you use. So there's some fancy machines you use here. Sometimes it's you do it in liquid, sometimes you do it on powder form. And then the time that it takes for verifying the C is being C really depends on what technology we're talking about. Sounds like that's
Andrey Ivankin
Correct.
Xiao Zhong
That's an interesting point, and frankly, that's a bottleneck I know for some of the subjects. As well, characterization takes a long time. Let's actually dive that deeper into what you are focusing on. I know maybe I'm wrong. You're focused on certain materials catalysis or the stuff like that. There's a partnership approach here. Whereas there's some other maybe other companies Lila and others there, they're more broadly focused on AI in science. I wonder why the choice of focusing on a certain material is it because you see, hey, you know what, not all material. Characterizations are equal. Sometimes they're really slow, others really fast. Let's pick the one that's fast, where also there's a degree of willingness to pay from customers. Is that kind of the approach? I'm curious to hear your thoughts.
Andrey Ivankin
Yeah, so where we start obviously you try to start with something easy and simple. So in that regard, we started with optical characterization and electrochemical characterization, which is technique as known as scanning droplet electrochemical cell. It's basically a small droplet where you run your chemical reaction and you can just raster that droplet across the chip and measure activity of a given catalyst, just within the droplet at every single point. So we started there because again those techniques are easier and faster to produce data. But that's not where we want to stop. We believe that our technology is very generalizable and that there are two steps to the endpoint where you have enough data you can generate it fast enough. First of all, it's interesting to see that field recognizes it. Recently we went to MRS conference and everyone is pushing characterization to its limit. And we want to actually expand atomic force microscopy. I know you know about it. We work with Bruker and Oxford on expanding and developing the characterization and application of that technique to megalibraries. And again, you will never have the throughput of optical microscopy, but you can run fast enough that it becomes very meaningful. With each instrument. And then atomic force microscopy I like to call it the Swiss knife of characterization because it can be operated in so many different modes. You can again look whether your materials are good magnets, whether they're good conductors, maybe you can look at the mechanical or thermal properties or piezoelectric properties. So all of that information becomes available. And then the structure is another thing that obviously is of huge interest because you only have material fully characterized when you have composition linked to synthesis, linked to structure and linked to properties that material produces. So right now we are using scanning electron microscopy with some diffraction techniques to characterize crystal phases and everything. But the biggest potential where we see the opportunity is for the STEM and EELS. The technology there is available to characterize tens of thousands of materials a day per instrument. If you think about what's available today humanity to date characterized and documented in the Inorganic Crystal Structure Database around two hundred fifty thousand different inorganic materials. When you combine megalibrary and 4D-STEM in automated fashion, you can surpass that in a matter of week with one instrument.
Xiao Zhong
Yeah, that's very interesting. And then I do wanna give a comment. So the Inorganic Crystal Structure Database here is behind a paywall, which I was reading something about this. So hey, there's Meta and also Google, they both have open source some databases there, and then they call them the ground truth. It's all computational results there. That's great. But there were maybe this as well, there were cases where it turns out, some of the structures they discovered are actually already in existence, developed by other people, just because there's this paywall. The I assume the train the researchers from those two companies were not able to get use those databases there. But anyway, just that's going off the topic. So basically what I'm hearing here is there's the two steps for generating a lot of data. There's A plus B equals C. You want to do that really quickly, which you solved. And then there's the step second step, which is C being really C, and then you are solving that as well. Now let's actually go back to the business model question there. So what is your so you're generating a lot of data? That data here is uniquely valuable in the sense that it's ground truth data, meaning it's actually done in the lab. And you can do this really quickly. And if all the hypothesis you mentioned here become true, you can quickly surpass the number of data points there that's ever synthesized in the lab by human altogether in a ridiculously short period of time. What do we do with the data? What's our business model?
Andrey Ivankin
So data by itself is not very useful. So but data can provide insights and the more data you have the more generalizable your insights become across materials and across the properties. Our thesis that most of our materials discovery companies failed because they couldn't generalize and what happened they had to actually focus on very specific vertical and very specific application and that doesn't scale very well so our premise is that by going at scale and broad across applications or properties, eventually we have the model that can get us answers across applications fast. And the beauty here that we don't only have that model, but we have the engine, real-world engine behind that allows us to quickly validate those predictions and refine the model, improve the model so we arrive to the answer in very few loops. In the earlier days we want to focus on few verticals again we start with catalysis and more broadly energy domain we are very interested in magnetics domain and semiconductor materials that there is a huge need to develop many different materials in there. The models that we partner with companies who have the materials problems because our data has two benefits it has it helps you find new materials faster it also provides you some guidance which materials are easier to scale than others and bring to market so basically partnering with players from the field, and providing intelligence and but help them make best materials, R&D decisions.
Xiao Zhong
Yeah, that's interesting. So what type of let's say personas are you talking to? I know this R&D there. What type of companies there are? And then let's say if I were from those categories there, how do I work with you there. So some of the people that's listening here may be potentially interested in working with you or are getting to know more about Mattiq.
Andrey Ivankin
Yeah. So we engage and obviously share our capabilities and make sure that we operate in the same domain, in terms of materials that are of interest and in terms of properties that are of interest. Then we define the problem statement what material you're trying to identify, what you're trying to solve for. And then basically we define the milestones and some budgets and capacity of the factory that are required to explore the universe that you're interested in and see if we can discover some new completely new materials and generate IP that's not out there. And at that point, once we generate this validated materials on chip. We pass along the formulations and you validate. In that sense, we don't think that we replace your R&D necessarily, but if you can do hundreds of thousands of experiments a year, we just give you the best chance that you will succeed with those experiments and we'll focus on the best materials.
Xiao Zhong
I've been tracking the materials informatics space, loosely speaking, materials informatics for a long time. And then suddenly we're seeing the few newcomers that have raised hundreds of millions of dollars there. That's suddenly at least an order of magnitude more than some of the other companies, including Mattiq. Why that's the case. What are these companies doing that are different?
Andrey Ivankin
I think one interesting shift in the past couple of years if you look at this GNoME model that Google DeepMind came up with in 2023, and then Microsoft came up with some models everyone was excited, but the reality is we are many years later and most of those predictions have never been validated. And I think the field is realizing if you look at DeepMind, they're building their own wet lab to run materials experiments. The field is realizing that we need experimental ground truth to train models and actually make something useful out of them. Not to say that the models are bad, but there is quite I like to say that materials AI today has a reality problem. The training data is mostly computational, most predictions never validated. We need to change that. And the field understands it, but the problem with that is that if you try to do it by automating the experimentation of with what we've been doing to date, you require a lot of CapEx, you require a lot of development. And development not only development because the whole field is not ready. Instrumentation vendors the day instruments are not ready for running in the loop self-driving lab. So you need to put a lot of effort into optimizing and developing APIs and so on and so forth. But also again to get to a meaningful scale you just need a lot of CapEx money if you're trying to do it via more traditional approach.
Xiao Zhong
Yeah, so totally agree. So but then someone has to do it. Otherwise, all those the results that are we're getting from those models are actually trained on computational data that we know there's an inaccuracies here and there. So is that kind of maybe why we're seeing companies raising hundreds of millions of dollars to actually do that? We still do simulation, we still do AI, but guess what? We're actually building self driving labs, and of course, the government here the Genesis Mission, it's one of the strongest emphasis here is on autonomous labs or self driving labs there as well. It's gonna take some time, but I personally do think, there's this gap that our materials models are not really living in this reality here, it's gonna be solved.
Andrey Ivankin
Again, going back to analogy with biology, we don't see megalibrary as alternative approach to what let's say Lila or Periodic Labs or Radical AI building. We operate at different scale. We work at small, nano scale, but the advantage from that we get is we can run much faster and more cost efficient than the other companies. At the same time what they do they're a little bit closer to the end application and format of materials how it's going to be synthesized and so on so forth. So that gives the next phase of the characterization. So we think about it as a funnel. [Unclear audio.]
Xiao Zhong
Yeah, interesting. Okay. So I last couple of questions they're really about you personally yourself. So you deep tech founder. You get a PhD, publish papers there, and then you transition several times there. Tell us more about that transition. What's been really fun, what's been challenging, or what's are what are the things that you feel like hey, it didn't want to do but you have to do. We talked about this before in different settings, but the tell us more, please.
Andrey Ivankin
That's an interesting question. So I always had that entrepreneur desire. So really want to do that. But to your point, when you become a founder, you think you will keep developing, especially coming from technical background. You think you will spend all of your time developing the most exciting technology by yourself writing code or working on hardware in the lab. The truth is as a founder, that there is a lot of things that you have to do and you have to worry about that there is fundraising stuff, there is hiring stuff. You need to interact with VCs and your supporters. Again, all of those are very important and necessary. Things and I don't necessarily say that I dislike it. What I'm saying is that the scope of your responsibilities expands very quickly from you being a grad student postdoc. In a way there is some similarity with professorship I think because again you have to worry about bringing students, getting funding. And making your department happy and everyone else happy. So yeah so the there are some more glamorous and less glamorous things that you will have to do but at then at the end of the day it's very fast paced it's very rewarding in terms of intellectually and if you an idea guy a girl and want to come up with and always come up with many ideas. That's the best thing to do.
Xiao Zhong
Yeah, so and then there's and I anticipate a lot of people that are listening here are someone let's say from a national lab or someone from a university, maybe a professor, maybe a postdoc, or maybe technical people here who have a lot of technologies there. They're I was actually talking to one of them a few weeks ago there he's very interested in commercializing his technology, but then in this should I leave the comfort zone here? Should I just become a CEO or should I be looking for a co-founder. So what you said there is really helpful. Ask yourself. So the question for example Andrey was talking about is that something that you're gonna be willing to do or even happily doing it, let's say hiring people, worrying about fundraising, other than just the developing technologies, or maybe if the answer some of them are no, you should be thinking about, partnering with someone else who has a sort of more of a business background as well. That's my personal opinion, maybe yeah.
Andrey Ivankin
I will say that I've seen some of the technical founders coming out from universities and not and remembering me when I just started. I was lucky when I started because my co-founder Chad Mirkin is serial entrepreneur by then and we had he had a lot of knowledge and we had a lot of resources. To get us started. But I see many people face a lot of challenges when they don't have all of them are first time entrepreneurs, they don't have the network and support that they need to start a company and that can be slow unproductive and painful so in that sense either finding experienced advisors or co-founders early on I think it's very helpful. You can save a lot of pain for yourself.
Xiao Zhong
Absolutely. So last question more for the company. Are you guys raising any money now?
Andrey Ivankin
Yeah, yeah, we have we currently raising series A to expand the scope of characterization, be and go into additional verticals and proof ourselves out in AI to build and demonstrate the first model strength on our data and benchmark against the other models out there.
Xiao Zhong
Yeah. So that's lots of insights. And then just in case you were wondering, they're raising money. So I'm gonna, put Andrey's LinkedIn link in with the with this podcast wherever we're posting it and also the company's website as well. Sounds like you're always welcoming a chat with potential partners say, who can be complimentary to the technology you're developing. Or maybe you can help them solve some of the R&D issues. Now you know where to find them. Again, this is AI Goes Deep. Really appreciate the chance to be speaking with Andrey. Thank you. Thank you again, Andrey.
Andrey Ivankin
Thanks a lot, Xiao.
Xiao Zhong
Okay, take care. Bye.
