Uncountable

[MUSIC PLAYING]

WILL TASHMAN: My name is Will Tashman. I'm one of the co-founders and the chief revenue officer at Uncountable. My background's in material science. I got my degree in that from MIT, and then I worked in engineering and design at Apple before starting this company. Our customers leverage Uncountable's platform to store, visualize, and analyze experimental data, and then we layer in machine learning algorithms on top of that to assist scientists when predicting how new formations might perform.

The idea of the platform is that everything now in 90% of companies we talk to is all done via Excel or lab notebooks. So you might have an experiment done in your own computer. I might have my own. And problem is that those experiments don't usually talk to each other. They don't learn from each other.

And so through the Uncountable platform, not only do we have a landscape that's flexible enough to handle data, but it's also a collaborative platform. I can look at my peers' projects were working in a similar chemistry area and quickly find something that's relevant to me and uncover new learnings that I might not have been able to find in the current regime of Excel and shared drives and very siloed work environments.

We view the platform as a way to drive ROI in a couple of different ways. One is sort of this micro-level workflow thing. So scientists are typically setting up experiments in one program then sending them out in another program and reviewing analysis and data and another program. We want to remove a lot of those the handoffs in the exchanges and formatting issues that scientists typically run in and make everything streamlined.

On top of that, we think about how we scientists find old, relevant experiments, whether they might be their own or someone in a different office, in a different country. That time could either be three to four hours, depending on what the current landscape is or it could be impossible for a current company today. And with the Uncountable platform, we shrink that down to three or four minutes usually.

At a higher level, we're always trying to help our scientists and help our customers move faster. So we want to help their scientists make better products and experiment more efficiently. And so there's a lot of tools built into the platform that allow them to either predict how new formations will perform or uncover these relationships that they might not have recognized because the analysis is all sort of streamlined and surfaced to them in an intuitive way.

[MUSIC PLAYING]

WILL TASHMAN: We work with a lot of different companies and industries. They range from polymer synthesis manufacturers who make polymers for compounds of rubber, to adhesive companies, to coatings companies, and then even into ceramics or consumer products, maybe like a sunscreen or a lip balm. They are all relatively different chemistries from a true fundamental standpoint, but the practical workflows, the practical data problems, and actually the statistical problems are quite similar. They're all working in these sort of multilayered, complex, formulated systems, where they have somewhere between 5 and 25 ingredients that go into a single product. They may measure 30 different things about that product. We need a way to quickly and easily synthesize that information so that it can be flexible enough to cover all those use cases, but also streamlined enough so that each step is fluid, and easy, and responsible for the scientists.

The platform itself is designed to be flexible enough so that any customer could go in and create it so it works and looks like it should look for their own work environment or work structure. But it's not necessarily like custom engineered per se. So we have people on our side who can configure it, where a configuration example might be, if you're an adhesive manufacturer, you might have a part A and a part B recipe. We need to set that up so it looks like that. Or you might have certain calculations you do from a stoichiometry perspective that might need to be inserted that are different because a rubber company may not have as many stoichiometry calculations needed. And so in that respect, a customer's deployment looks unique because it is unique to them. But from a software standpoint, the tools and the flexibility are all there from a configuration.

WILL TASHMAN: As with all machine learning applications, the more data the better. But we always think about this as how do we help scientists and companies get better, not just be perfect? The idea that you're going to have some magical system tomorrow that can pull stuff out of a hat just isn't really practical in this day and age. We always think about this in terms of can we make scientists 10% more efficient, 20% more efficient, 50% more efficient? And the way to do that is to start to build out this infrastructure of data that's correctly labeled, correctly structured, and then on top of that, we can layer in these algorithms that are actually tuned for small data problems. Big data doesn't really exist in materials or chemicals, right? There's the idea of deep learning. We're not quite there yet within a company, in terms of how much data does an actual chemical company collect within a year with regards to formulations? And so our models are specifically tuned to these smaller data problems. So they're more flexible. They handle noise better than some of the bigger data techniques.

On top of that view the AI as helping to fill in a lot of the gaps in a scientist's workflow. So an automotive customer may require testing at 1,000 or 1,500 hours, but if we can use the model to actually predict that 1,000 hour performance or 1,500 hour performance using only the 100 hour or the 150 hour data, we can save weeks at a time in these testing cycles. And then they can still go do that final validation step before they go send to a customer. To imagine a world where automotive and aerospace companies aren't requiring 2,000 hour, 3,000 hour tests for a lot of their products probably won't be in the near future. And that's from a liability standpoint. I totally understand where they're coming from. But from an iterative standpoint, so what am adhesive manufacturer or paint manufacturer may do in their development, I think there's big opportunities to change there.

We oftentimes will sort of highlight these relationships, or the strong relationships between different condition sets of a test, whether it's maybe different substrates, different aging times, or temperatures, or maybe aging mediums, between these tests where they may be technically different, but actually the chemicals or the materials perform relatively similarly or in a correlated fashion. And so if we can help scientists just run less types of tests or shorter tests, we think there's a good opportunity there to make them go faster and think about this in a different way, rather than they have to go and wait for a six week test to run fully.

WILL TASHMAN: Data security is our number one priority as a company. We have to be an extremely secure, extremely private company to be able to work with all of our customers. Some of our customers are direct competitors of each other, and for them to trust us with their core critical formulation data, it's a big trust factor to have.

And so we do that by having industry leading and best practices in the security field. So we go through third party audits. We use the industry leaders in services who use AWS. We have sort of more technical deployments that we can do and longer relationships with our customers.

And internally we have to follow extremely strict regulations and rules for our own company, where everything's on a need to know basis. So if you aren't working with that customer directly, there's no reason for you to have access to that. We are relatively siloed in that respect. So our own individual engineers may not be able to talk about exactly what they do. I came from Apple, I'm well used to this siloed environment, and I've tried to apply a lot of those practices here.

When you're developing a new material, whether it's a glue or a tire or a paint, there's several steps in that process, even in the lab skills, that go into it. You need a system and a platform that's flexible enough to handle that. So you get ingredients from suppliers. Ingredients have different batches. Ingredients have different attributes. So you get a polymer, it's going to have a TG, it's going to have a molecular weight. You get a filler, it's going to have a particle size or a surface treatment.

Those things, those attributes, are what scientists are using to think about hypotheses as to why this ingredient will help or hurt this product. We need to be able to incorporate those. A scientist may then go into a lab and mix up six formulations at a time, and then measure three different things about each formulation three different times itself. So need the ability to handle replicates.

We want the ability to take formulations that you make, and then quickly figure out whether someone's made something similar in a lab in Munich or in a lab in Tokyo. We need the ability to have flexibility for process parameters, or multi-step phases where you mix four ingredients and you mix six ingredients, and you mix these final two curatives.

All these different sort of configurations bring complexity to a software product, but it's a complexity that our customers deal with day to day. And so we need to be able to have a solution that's comprehensive enough and is easy enough to use so that they quickly figure out, OK, here's where my replicates go, here's where my part A, here's where my part B is going to go, and here's where my calculations are going to come from. And that's a lot.

And then beyond that, you have all of your analysis and you have your different batches for day to day. But you need to have this multilayered system so that people can trace back an ingredient which might be technically another formulation, or a master batch, or a combination of ingredients in a multi-step world or a multi-step process that our customers live in.

[MUSIC PLAYING]

WILL TASHMAN: A couple of things we think about in terms of our company principles is one as being a customer-first company. We spend a lot of time with our customers trying to figure out exactly what their pain points are, either in their current workflows or when starting to use our platform. We feel that, if we can make a scientist's life 10%, 20% easier, we think that we're going to start to earn their trust and initiate this cultural change that needs to happen amongst these R&D teams.

We have a software and ML focus first as a company. So I have a material science background, but I don't do anything technically. Our other two co-founders are both MIT and Stanford grads in machine learning and computer science, and they're really the experts there.

And so we've created a software company for material science rather than being chemists who go and hire software engineers. And so that leads a lot of our guiding principles and how we operate and how we develop products for our customers. And I think we've also narrowed down to a somewhat niche space, in that we don't work with every single science company in the world.

We focus on companies that are formulating or synthesizing these complex products, and are they going to measure a lot of things about them? Now, that goes from adhesives, to ceramics, to consumer products, but we don't work with every single large company that might have a chemical focus, per se. Maybe in 5 to 10 years we will, but we're focusing now where our customers live and how they work and want to make sure those early customers are true partners for us, and we can demonstrate success with them.

We opened up an office in Munich, Germany, at the beginning of this year. We plan to grow out that office, as well as our headquarter office, in San Francisco, substantially over the next two to three years. We probably want to open up an office in Japan. We have a couple of customers over there, and we want to make sure we can support our Japanese customers the same way we can support our North American and European customers.

[MUSIC PLAYING]

[MUSIC PLAYING]

WILL TASHMAN: After I graduated MIT, I was sort of vaguely aware of MIT's presence in the startup community, but it wasn't until I started Uncountable with my other two co-founders that I recognized the true benefit of the MIT network. STEX 25 has provided a lot of awesome opportunities to present at the small conferences or larger conferences in different chemistry areas or different locations. They also have really, really strong connections in larger industries and larger companies that we want to work with.

And coming with the name brand and the credibility of MIT when many of these chemical leaders come from MIT or know someone or spent their time here for their PhD, I think it adds a layer of credibility to us and allows us to skip one or two steps in this long prospective customer journey that we go through with the these chemical companies. Along with our customers being large international companies, they're also typically very well established old companies where they may have been making products before World War II.

Obviously, we're a very young company, and we come with a Silicon Valley mindset to this industry. And so we want to move fast, and all of our customers are very accustomed to moving slower. And we recognize that.

And so we want to be acting as an agent of change for our customers so that our executive leaders can go on point to us and say, here's something that we're trying to do that's innovative. But we also recognize the pace of these companies, and we're not going to have a overnight shift in how these scientists who have been working in tires for 30 years operate. We want to start slow, build trust, and sort of use a snowball type of effect to move faster and faster as we get going.

But there are a crawl, walk, run procedure here that has to happen so that we can prove point or demonstrate proof points within these smaller teams, demonstrate scalability within the small or medium-size teams, and then demonstrate sort of this full-fledged effort at the large scale, and that process doesn't happen overnight. It doesn't happen within a year. It probably doesn't even happen in two years. So we're always thinking about ways that we can paint the right picture and demonstrate that for these bigger customers is what it could their program or what could their R&D team look like and how could they operate in three to five years when they have had the experience and the proof points in the structure of working with Uncountable.

[MUSIC PLAYING]

WILL TASHMAN: Uncountable provides R&D organizations with a data platform that enables collaboration, knowledge management, and faster development times for products. Scientists can use the platform to store, visualize, and analyze critical experiment data, from raw ingredients and different suppliers, to experimental results. We then layer in the advanced machine learning models that enable scientists to predict how new formations perform allowing them to move even faster. Our goal as a company is to be a true partner for these R&D teams in their digitalization effort, and as they move to a more effective and efficient environment for creating products.