Bioptimus raises $41M to develop a ‘GPT for biology’

A fledgling French AI startup has raised $41 million to develop a foundational AI model for biology.

Just as OpenAI’s ChatGPT has taken the world by storm for a super-smart generative AI tool capable of natural language conversation in text form, Bioptimus is taking that basic concept but training its model specifically for downstream biological applications — something that comes with its own unique set of challenges, given that the required clinical training data is inclined to be sensitive, and not publicly available.

Bioptimus co-founder and CEO Jean-Philippe Vert says the company is looking to develop a greater understanding of biology by learning how it works from raw data spanning molecules to entire organisms. This, he says, will enable scientists and researchers to simulate the biological world to “predict disease outcomes” and develop more effective treatments. And it’s this simulation that Vert says makes its technology a little akin to what ChatGPT’s underlying model is all about.

“Beyond pharmaceuticals, this model will unlock limitless possibilities across many other industries, driving biological discoveries in ways we are only beginning to imagine,” Vert said in a statement. “Essentially, it’s like the GPT of biology—but instead of generating text, we’re simulating biology.”

France has emerged as something of a hotbed for AI startups, with generative AI companies across the country securing the lion’s share of funding as of last year. Mega funding rounds include Mistral AI’s $640 million tranche, “H” securing $220 million, and Hugging Face closing a $235 million investment — all in the past 18 months.

Bioptimus, for its part, was only founded last year, but already it raised a $35 million seed round. That it has now raised a grant total of $76 million, less than a year from its foundation, is testament not only to the current AI hype, but also the backgrounds of Bioptimus’s six co-founders. Chief technology officer (CTO) Rodolphe Jenatton, for instance, was previously a senior research scientist at Amazon and Google. Vert, meanwhile, is not only co-founder and CEO at Bioptimus, but he’s also chief R&D officer at Owkin, a French unicorn and yet another AI-infused biotech startup with backers including GV.

This dual role hints at Bioptimus’s origins. Owkin leverages AI and machine learning to accelerate drug discovery, and has built a swathe of partnerships with top biopharmaceutical companies. As part of this work, Owkin has also amassed a ton of multimodal patient data, which is what Bioptimus will using to train its foundational model.

So, rather than creating a tangential unit inside Owkin that was focused on foundational models, it simply made more sense to create a separate entity.

“Building biology [foundational models] is not a part of Owkin’s roadmap, but Owkin supports and is keen to partner with a company like Bioptimus,” Vert explained to TechCrunch in an interview last year. “Training very large-scale [foundational models] requires important resources in terms of data volume, computing power and breadth of data modalities that are easier to unlock as a specific entity. As a ‘pure player’ in foundational models, Bioptimus is better set up to do this.”

In the intervening months, Bioptimus launched H-Optimus-0, an open source foundation model for pathology, which was trained on millions of images to help in the research and diagnoses of diseases, such as cancer.

Bioptimus’s latest cash injection was led by U.S. venture capital firm Cathay Innovation, with participation from Sofinnova Partners, Bpifrance, Andera Partners, Hitachi Ventures, Boom Capital Ventures, Pomifer Capital, Sunrise, and several angel investors.

The company said it would use the funding to bolster its AI platform with a more diverse array of data sources covering broader therapeutic areas, while it will also look to build further partnerships with the pharmaceutical and biotech sector.