AgFunder leads $5m seed round for ML-powered Atinary Technologies to turbocharge R&D process

[Disclosure: AgFunder is the parent company of AgFunderNews.]

Atinary Technologies, a Swiss-American startup deploying machine learning (ML) to accelerate the materials discovery process, has raised a $5 million seed round led by AgFunder and supported by Cherubic Ventures.

The investment will help the company — which has operations in Lausanne and Silicon Valley — expand its team, build partnerships, and deploy its ‘no-code’ Self-Driving Labs (SDLabs) platform at scale.

“The current state of materials and molecules discovery is slow and costly,” said Manuel Gonzalez, partner at AgFunder. “Atinary ‘s innovative approach acts as a time machine, dramatically speeding up the process and unlocking the true potential of R&D.”

Materials discovery… accelerated

Atinary Technologies was founded in 2019 by economist-turned-entrepreneur Dr. Hermann Tribukait, and chemist and computer scientist Dr. Loïc Roch. Prior to this round, the company had raised just over $2 million in pre-seed funding from small private investors and almost $3 million in grants to build a cloud-based machine learning platform that it claims can dramatically speed up the R&D process for discovering breakthrough materials and molecules.

While the platform is high-tech, users do not need to be software engineers to use it and can integrate it into existing workflows with minimal onboarding, claimed Roch.

“Our SDLabs platform offers optimization and discovery solutions that surpass traditional trial-and-error methods such as one-factor-at-a-time (OFAT), design of experiments (DoE), and high-throughput experimentation (HTE).”

SDLabs could potentially be used in biotech, food, chemicals, energy storage, and climate tech, Tribukait told AgFunderNews. However, the sweet spot applications are in chemical formulation, catalysis, and synthesis, where the SDLabs software platform allows companies and R&D labs to optimize experiment planning, orchestrate workflows, and hit targets potentially 10 to 100 times faster.

‘Imagine you’re baking a cake…’

The platform is particularly well-suited to solving multi-objective optimization problems with multiple parameters, including categorical variables and non-linear constraints, added Tribukait.

Imagine you’re baking a cake, but you’ve never baked one before, he said. What ingredients do you use, in what quantities and order, how long do you bake it for, and at what temperature?

“So with SDLabs, you run your first experiment, or [to continue the metaphor] you bake your first cake and you taste it,” said Tribukait. “That’s the first iteration. And then the algorithms capture the results. And then we retrain the algorithms with the latest data. These are general-purpose Bayesian optimization algorithms, so you retrain and update your algorithms after each iteration, and then the algorithms refine their prediction and recommend the next set of conditions.

“The more variables you have, the number of combinations is potentially humongous, and if you’re using traditional processes, it’s literally like trying to find a needle in a haystack of millions of combinations. You’re just guessing.”

Digitizing the R&D process

He added: “What Loïc has essentially built is a software platform that allows any chemist to deploy machine learning by just clicking a button. So it’s a no-code platform in the cloud, meaning you can access it anywhere, as long as you have an internet connection, and you don’t need to be able to code.”

In a nutshell, he said, “SDLabs allows you to deploy machine learning to optimize existing materials or discover new ones in the broad sense of mixing different chemicals to come up with a new chemical. That could be a drug, it could be electrolytes for batteries, materials for solar cells, or a new formulation for food.

“It also helps digitize the R&D process, where many companies are still stuck with Excel spreadsheets. After a two-hour onboarding session, you can start optimizing your experiment in your existing workflows.”

Made by a chemist for chemists

Unlike a lot of software deployed in the industry, he explained, “One of the distinguishing aspects of SDLabs is that this is a product that was from the idea through to execution developed by a chemist. So that makes it a very user-friendly, very intuitive product to use for chemists as it speaks the same language. We didn’t go to software developers and say, ‘Can you develop this for us?’

“And it’s solving a real problem, which is that chemical R&D, but R&D in general, is just too slow. It’s painful. It’s repetitive and inefficient.”

If your lab is fairly low-tech, he said, “You don’t need automation or robots to use this, but if you want to add automation to your workflow, one of the distinguishing aspects of our platform is that it can directly seamlessly integrate or talk to those automated platforms. Now we don’t offer robots, but we do have close collaborations with robotics companies.”

Progress to date

So what progress has Atinary Technologies made to date?

“We signed our first client in late 2019,” said Tribukait, “Most of our revenues to date have been from paid pilot projects, but we have since moved to our first commercial agreements (annual subscriptions) so we’re at that stage of starting to scale. So with this investment, it’s all about scaling the team, deploying the technology, and building partnerships.”

Several partners wish to remain confidential, he says, but companies that have publicly mentioned working with Atinary include flavors and Fragrances giant Firmenich, which merged with DSM earlier this year.

While there are now several other companies developing AI-powered platforms to accelerate the R&D process, he said, “We have an outstanding multidisciplinary team spanning software development, AI and machine learning, robotics, computer science, and chemistry.

“The other thing that tends to be under-appreciated is our user-friendliness, the ease with which this can integrate with automation and robots, and finally our proprietary machine learning algorithms, although we also offer open source algorithms.”

Time machine (learning)

Gonzalez at AgFunder added: “Recently, I was watching a keynote from NYU professor, Scott Galloway, where he was essentially saying that most $100 billion companies are time machines. When I met the founders of Atinary, I felt that I was right in front of a time machine.

“This company, with a no-code ML platform, expedites R&D, optimization, and discovery in chemistry by a staggering 10X to 100X. This created my conviction that Atinary is a game-changer for R&D.”