Dr Loïc Roch works at the intersection of chemistry, artificial intelligence (AI) and computing. He explains how he got to where he is today – and how AI can make research and development (R&D) in the life sciences and materials sciences more effective.

Tell me a bit about your background and how you got to where you are today.

I’m Swiss and my background is in chemistry and computing. I did a PhD in quantum chemistry in Switzerland and China before deciding to move to computer science. I developed Atinary’s underlying technology while working on AI and software design for my postdoc at Harvard. At a workshop in Mexico in 2017, I met my co-founder and we decided to launch a start-up in 2019. We had complementary skills as he’s from a business and economics background. After raising funds for Atinary in the US, we came back to Lausanne and built a team here – so after spending time in lots of different places, I’m now back in Switzerland!

Atinary is a Swiss-American company. What are the benefits of being based in both countries?

There are lots of advantages. Starting a company in the US is extremely quick – it can be up and running in a matter of days. The US is more opportunity-oriented and less risk-averse than Europe, so we managed to gain traction and raise capital quickly. We do strategy and business development there, together with our business advisors, while the rest of the team is based at Biopôle. In Switzerland, we benefit from a rich ecosystem that combines leading companies in chemistry, pharma and life sciences with top universities and talent. And at Biopôle there’s a real spirit of collaboration and lots of opportunities to work in partnership with others, which is a huge advantage.

What does your work at Atinary involve?

We merge high-performance computing, materials sciences and AI to fast-track R&D. Companies can deploy our AI technology seamlessly in their R&D processes and existing workflows. We have a product available online that R&D labs can connect to in order to accelerate their research. If they have automation available, we can make R&D autonomous, running 24/7 and leveraging and augmenting the work of the scientists in the lab.

How exactly does this work?

Imagine that you have a kitchen with lots of ingredients. I ask you to bake a cake, but you’ve never baked a cake before. What do you put in? What quantities do you use? How long do you bake the cake for? At what temperature? It would be tricky to get it right.

Now imagine an algorithm that tells you: put these ingredients in and bake the cake for X amount of time. You taste it and can start to rate your cake; you feed this back to the machine, which learns to make it better.

This is what our algorithms do, but in the context of R&D. Machine learning algorithms send the instructions to run experiments and get the output data back. For instance, for a life sciences project, our algorithms help to find the right chemicals to achieve optimal cell growth. They do this by suggesting parameters, such as how much serum should be used, what temperature to use, for how long, etc. Researchers feed the results back to the algorithm to train it, and it then gets better and better for their particular project. This approach is particularly useful in projects involving growing cells with lots of parameters involved, like stem cells.

What makes an algorithm better than a human being at making these decisions?

Humans are good at thinking in three dimensions – perhaps four at a push – but when you need to navigate 30 parameters, it’s extremely hard. We simply cannot process so many parameters at once, whereas our algorithms can analyse up to a hundred different parameters simultaneously.

How will AI lead to innovations in the field of life sciences?

Discovering new materials – such as biomaterials or pharmaceutical substances – is a very slow and expensive process. It can take many years and tens or hundreds of millions of dollars. This is the problem that we’re tackling at Atinary. Our technology fast-tracks R&D and makes it cheaper. Using machine learning and our cloud-based software platform, we have the power to process the data and suggest experiments that will lead to those discoveries faster. The impact on life sciences, materials sciences and chemistry will be huge.

Could you give us an example of how you’re doing this at Biopôle?

In order to demonstrate how our machine learning algorithms and software platform can accelerate R&D, we designed and built a 3D-printed self-driving lab prototype. With this testing lab, we showcase how our technology can orchestrate and execute chemistry experiments with complete autonomy, driven by our machine learning algorithms. We chose to do this with a simplified formulation experiment that is relevant to addressing the challenge of clean energy storage. Specifically, at Biopôle’s StartLab, we use our machine learning and data-driven approach to optimise bio-inspired electrolytes for next-generation batteries, which will help in the quest to develop safe and affordable clean energy storage or battery storage.

What are your three top tips for aspiring entrepreneurs?

1. Surround yourself with the right people. This is important because there are billions of things to do when you start your own company and you’ll need support. You’ll want trusted advisors or mentors who can help on various fronts.
2. Be prepared to jump on a rollercoaster with lots of ups and downs. Take the time to enjoy every single win. Even if it’s a small win, take a break and recognise that it is a win, because there will also be failures along the way.
3. Enjoy what you’ve built because it’s a wonderful feeling. Every day when I go to the office and work with the team, it’s amazing to see what we’ve created from zero.

Do you sometimes get the urge to get away from tech and data?

Absolutely. It’s important to find the right work–life balance. I love being in nature, mountaineering and trail running, so that’s what I do to get away from it all and clear my head.

Complementary content on this topic