London biotech Helical has raised $10 million to build a governed, reproducible layer that sits on top of the 350-plus AI models trained on DNA, proteins and cells that pharma R&D teams now use to predict how drugs will behave before running a physical experiment. The seed round was led by Swiss investor redalpine, with Gradient, BoxGroup and Frst joining. Cohere chief executive Aidan Gomez, Hugging Face chief executive Clement Delangue and footballer Mario Goetze put in personal cheques.
The company is already live inside multiple top-20 pharma companies, including a public collaboration with Pfizer on blood-based safety biomarkers. Helical's pitch is that the bottleneck in AI drug discovery is no longer model quality. It is that the output of any one model is almost impossible for a regulated R&D organisation to trust, reproduce or defend.
Enjoying this article?
Subscribe to our newsletter and never miss a story. No spam, ever.
Why pharma stopped building its own AI stack
2026 has been the year pharma capitulated. Eli Lilly signed with Chai. Novartis extended its tie-up with Isomorphic Labs to a deal worth up to $3 billion. Bayer bought into Cradle. GSK paid Noetik $50 million upfront. Each of those deals buys access to a specific proprietary model or pipeline. None of them solves the problem that an average big-pharma machine-learning team is now tracking dozens of open bio foundation models, from scGPT to Evo to ESM to AlphaFold3, with no shared environment to run them in.
That sprawl is the opening Helical is walking into. Scientists and ML engineers end up rebuilding the same notebooks for every new hypothesis, and results that look promising in one lab rarely survive the trip to another. The company's bet is that the pharma AI winners will not be the groups with the best single model. They will be the groups that make every model usable across a company.
The product: two surfaces, one system
Helical splits its platform into two halves that sit on the same data and the same results. The Virtual Lab is for biologists and translational scientists, who can run hypotheses against any supported foundation model without touching code. The Model Factory is for ML engineers and data scientists, who can register, compare and deploy new architectures as they appear. Both sides see the same experiments, so a biologist's result and an engineer's tuning work feed the same evidence trail.
The standard complaint inside pharma ML teams is that computational predictions and bench work live in separate tools, and the handover loses context every time. Helical is trying to collapse that handover into a single workflow that survives audit.
"The models alone don't discover drugs. The system does," co-founder Rick Schneider said in a statement. "Pharma teams need a system that turns foundation models into workflows scientists can run, validate and defend. We built Helical to make in-silico science reproducible at pharma scale, so teams can go from hypothesis to decision in days instead of months."
The Hugging Face parallel is not an accident
Hugging Face built a company on the insight that once foundation models proliferate, the value shifts to whoever hosts, governs and standardises them. Helical is making the same bet in biology, a field where reproducibility stakes are much higher because the downstream decision is which compound a pharma company spends the next decade and two billion dollars on.
Schneider, Maxime Allard and Mathieu Klop founded Helical in early 2024. The three met at school. Schneider built enterprise software at Amazon and then helped Celonis scale into France and Japan. Allard led data science teams at IBM before a PhD at Imperial on reinforcement learning and neuroevolution for robotics. Klop is a cardiologist and genomics researcher.
Helical co-founders Rick Schneider, Maxime Allard and Mathieu Klop.
What redalpine is really buying
redalpine raised a $200 million fund last year, and half its investment team are scientists. Its bio portfolio leans toward platform plays and its AI bets include Mistral and Lakera. Helical is a clean expression of both: software margins on top of biology, and infrastructure rather than another drug pipeline.
"We are at a unique point in time where biological foundation models and general language reasoning models are converging," said Daniel Graf, general partner at redalpine. "We backed Helical because we strongly believe they have what it takes to build the pharma AI orchestration platform that will drive this transition from siloed AI models to integrated virtual AI labs."
Global pharma R&D spending is above $300 billion a year. The cost of bringing a drug to market, counting failures, is roughly $2.6 billion. More than 90 percent of candidates entering clinical trials fail. Helical is not promising to move those numbers on its own. It is promising that the computational step upstream of the wet lab stops being a source of noise.
What the $10 million buys
The plan from here is to go wider inside existing pharma clients, add more top-20 organisations, and grow the deployed-science engineering team that sits alongside customers. Helical's moat is less about model IP and more about the evidence layer that compounds every time a pharma team runs another experiment inside the platform.
The $10 million seed is small next to the Lilly and Novartis numbers. Helical is betting it does not need to match them to own the governance layer underneath.
FAQ
What is in-silico drug discovery?
In-silico drug discovery uses computational models instead of physical lab experiments to test biological hypotheses, predict how a molecule will behave, or identify promising drug targets. It does not replace the wet lab, but it lets scientists rule out bad ideas and prioritise good ones before committing the time and cost of bench work.
What is a bio foundation model?
A bio foundation model is a large machine-learning model trained on biological data such as DNA sequences, protein structures or single-cell measurements, in the same way a language model is trained on text. Examples include scGPT, Evo, ESM and AlphaFold3. They can be fine-tuned for specific tasks like predicting protein function or identifying drug targets.
Who are the leaders in AI drug discovery?
The field has two camps. Vertically integrated platforms such as Isomorphic Labs, Recursion, Insitro and Chai Discovery build their own models and, in some cases, their own drug pipelines. Infrastructure and application-layer companies such as Helical, Cradle and parts of Ginkgo sell tools that let pharma teams use foundation models inside their own R&D processes.
What does Helical actually sell to pharma?
Helical sells a software platform, not a drug pipeline. Pharma teams use it to run experiments against bio foundation models in a governed environment where results can be reproduced, compared across models, and audited. Biologists work through the Virtual Lab interface. ML engineers register and tune models through the Model Factory. Both sides share the same data and results.
