This technical information has been contributed by
Citrine Informatics

Artificial Intelligence Algorithms are Key to Development and Selection of New Advanced Materials

Mark Shortt
Editorial Director
Design-2-Part Magazine

Citrine’s data analytics platform is said to reduce the cost, time, and risk associated with materials R&D and manufacturing

Greg Mulholland is looking for a few early partners—large and small companies—who can help demonstrate the value of the materials data analytics platform that he’s presenting. Mulholland is co-founder and chief operating officer of Citrine Informatics, a two-year-old startup based in Redwood City, California, and he is speaking to an estimated 300-plus technology R&D, manufacturing management, and investment professionals at TechConnect World Innovation, a conference and exposition held June 14-17, 2015, in Washington, D.C.
 
Given 10 minutes to make his presentation, Mulholland describes a project in which Citrine’s predictive artificial intelligence system helped materials scientists at the University of California at Santa Barbara uncover what he called “a totally new class of thermoelectric materials.”

“The reaction that we got from the materials scientists at Santa Barbara was surprising,” he said. “They said that never in a million years would they have thought of this materials type to exhibit thermoelectric properties. But we have a partnership with them, and our models indicated that this would be a good thermoelectric, and indeed it was a thermoelectric material. Our models allowed us to reach outside of the normal thinking for new materials discovery.”

Advanced materials—whether they’re used in aerospace parts, advanced batteries, solar cells, or even polymer coatings—have long been developed by scientists who rely on their own intuition to develop those materials. But they are sufficiently complicated these days that no single human can hold the entirety of materials data in his or her own head anymore. That’s where Citrine’s data analytics platform comes in. “We’re basically taking a Big Data approach to chemicals and materials discovery and optimization,” Mulholland said in a follow-up interview at the TechConnect Expo.

The platform, which Mulholland calls “a tailored artificial intelligence engine for materials,” aggregates large-scale materials data by extracting them from journal articles, academic peer reviews, existing databases, and a customer’s own data. “There’s a huge amount of materials data out there in the public sphere,” said Mulholland in the interview. “Even if you go back in to the Cold War era, a lot of good materials research was done back then, so we go back and scrape that data. But we also use customer data. If you’re a large materials company and have 50 years of materials research, we actually will simulate that into our database for your use, and we’ll build models on top of your historical data. And so, as a materials company, you’re taking advantage of the last 50 years of knowledge that you’ve generated.”

Citrine’s platform then deploys highly tuned artificial intelligence algorithms to predict the performance of new advanced materials and identify those materials for specific applications in aerospace, advanced batteries, solar cells, polymer coatings, and the like. The algorithms are built on state-of-the-art machine learning algorithms, which are modified by Citrine to take advantage of physical phenomena present in materials and chemicals. “We’re specialists in designing materials-specific machine learning algorithms,” Mulholland said.

What types of insights can a company glean from all of that data? Mulholland said there are a number of possibilities, beginning with the ability to fill gaps in existing data. If you don’t know something about a particular material, Citrine can help you understand what that property might be. Perhaps even more exciting, he said, the analytics platform can help identify materials outside of the normal performance envelope that exists today.

“You might be looking for a lightweight alloy for a car. We can identify materials that are high strength, light weight, and take advantage of the data and the processing capabilities that you have in house already. The core for us is identifying materials that are able to be brought to market quickly for our customers. Discovery is really important, but more important is the ability to take that discovery and actually deploy it into a product,” Mulholland said.

Citrine’s artificial intelligence systems have already been demonstrated to work with phosphors, as well as thermoelectrics, and can be applied to a broad range of materials. Mulholland credits this to the strong collective skill of the entire team, which currently includes four others. Jordan O’Meara, the vice president of engineering, earned his B.S. and M.S. in computer science from North Carolina State University and previously worked for eight years at Red Hat, the developer of open-source software. Two colleagues who co-founded Citrine with Mulholland—Kyle Michel and Bryce Meredig—are materials scientists by training, like Mulholland, and have strong backgrounds (Ph.D.’s) in computational materials science.
 

“The Citrine team came together around the vision that future breakthroughs in the materials industry will emerge from large-scale data analysis and data-driven decision-making. The founders of Citrine hail from diverse materials backgrounds, but share a common belief: By combining data from across the materials community and applying state-of-the-art modeling algorithms to those data, software will directly accelerate materials innovation by providing unexpected insights and predictive guidance.” Source: Citrine Informatics (www.citrine.io/#citrineinformatics)

“I come from an experimental materials background, so we bring the best of both computation and experimental logic together, and that allows us to work with very, very large data sets,” said Mulholland. “If you ask me if I can find a totally new class of superconductors that is superconducting at room temperature, the answer is ‘absolutely not.’ But if you ask me if I can find existing phenomena in new materials classes, or that extend the envelope of known existing phenomena, or at least demonstrate the existing phenomena, that is exactly up our alley. So [that includes] things like identifying new polymer blends, or identifying new hard glasses, even identifying new catalysts. These are all areas where we’re very interested in working and we are looking for partnerships in.”

Citrine Informatics is eyeing two types of core companies as potential customers: producers of advanced materials—companies like Dow, DuPont, Corning, and Alcoa—and consumers of those materials. Consumers of advanced materials include large companies, like Boeing, Lockheed, Apple, or Samsung, that rely on advanced materials to enable their products. Although they may not be specifically focused on developing materials themselves, they use advanced materials as a differentiator, Mulholland said.

Citrine’s customer base also includes—to the extent that they need specific materials information—parts makers, contract manufacturers, and suppliers to product manufacturers and OEMs. “We see ourselves as the authoritative source for materials information, and so any company that relies on that fundamental materials information, we see as our potential customer,” Mulholland said.

At TechConnect, Mulholland said he was most interested in taking advantage of the opportunities to connect with some of the bigger companies. When reaching out to larger organizations, he said, it can be hard to find the right person who has both an open mind about new technologies like Citrine’s, and the ability to make a decision to bring a company in. The great thing about TechConnect, he said, is that “there’s a huge population of those people here.”

“We’re also looking to find partners that are smaller—companies that are agile, can move quickly, and can help us prove out our technology in a more public way and really show some early wins for our company.”

What’s in Your Data? After catching Greg Mulholland’s presentation on Monday, June 15, at the TechConnect World Innovation Conference, we caught up with him the next day at Citrine Informatics’ booth on the expo floor. Here’s more of what the co-founder and chief operating officer of Citrine Informatics told us about his company’s mission to help companies unlock hidden value from their materials data: D2P: What would you say is the biggest challenge in talking to people about the need for a data analytics platform for advanced materials? Greg Mulholland: I think the biggest challenge for us is that materials have been developed in the same way for a really long time, and I think everyone knows that data is the future, but I don’t know that everyone always knows what that means. And I think even we are discovering what that means. And so telling the story and explaining the product in a way that shows that it’s an enabling technology, not a threatening technology, as well as something that can be easily integrated into current development and production lifecycles, is something that is a challenging thing for us to do. That sort of communication is the hardest part.   D2P: How do you go about implementing this analytical platform with a company? GM: What we do is we make it as easy as possible. We take a couple of steps. First, we ask to work with their structured data, so, if they have Excel tables or databases of their own, we are able to bring those in very quickly. We also ask if they have things like technical reports and fields they’re interested in, old PowerPoint presentations that were given at technical meetings, and we actually have tools that go in and pull data from those reports and allow us to add them to our database. And then we work with a handful of scientists, or even one contact there, to identify the properties of interest that they’re focused on. We build models, or allow them to build models, that will predict these materials properties. And we deploy it, usually over the cloud, for them to triage experimental ideas or group force test thousands or millions of materials very quickly. D2P: How do you see the market for your product? GM: It’s something we see growing very quickly. When we started, it felt like it was potentially big, but it started as a niche. I think now, we’ve started to understand that there are a huge number of companies that are really hungry for data, and to use their data better. Executives at huge multi-national companies say ‘We have 100 years of research data that we’re not using anymore. How do we make value from that?’ And Citrine is the way they make value from it. D2P: Is that a common statement that you hear, that ‘we have so much data, but we don’t know what to do with it’, or ‘we don’t know how to access the relevant data?’ GM: Oh, it’s a huge problem. And I’ll tell you what’s even more striking than that. I can’t tell you how many people—the CEO or CTO of company X, and company X is many, many companies—have said, ‘We need to start putting in a data management system, we need to start making use of our data. Please go do that.’ And, all too often, in the materials industry, there’s really nowhere to go, and we see ourselves as the answer to that question. So it’s nice to know that there are scientists who want to use us to make their research better, and executives who are willing to both pay and get behind the implementation of a powerful system like ours. D2P: Do you see this as sort of being the “wind at your back,’ in a sense—the fact that data analytics has caught hold in a lot of other industries, like even in sports, for example? Is that something that you see helping your company? GM: Absolutely. I mean, I think the materials and chemicals industry is full of innovators, but often moves slowly because of regulation, because of risk. We’re fortunate to be in a position where the wind is at our back, from a data standpoint, and we are blazing new ground in the materials industry. So, it’s kind of an ideal situation for us. We have the challenge of education, but we also have the excitement of blazing new trails.