Pomegra
Paris-based Rivercell exits stealth with a €22M seed led by HV to build an AI virtual cell for drug discovery, trained on its own automated wet lab data.
Key Takeaways
- Rivercell raised a €22M (about $25M) seed round, led by HV, and disclosed no valuation.
- HCVC, Alven and Bpifrance Digital Venture joined; CEO Yann Fleureau founded the company in summer 2025.
- The money funds an expanded Paris wet lab and the launch of the AI Virtual Cell programme.
Lead
Rivercell, a Paris biotechnology and AI company founded in summer 2025, came out of stealth in early October 2026 with a €22M seed round, roughly $25M. HV led the round. HCVC, Alven and Bpifrance Digital Venture also participated. Rivercell did not disclose a valuation. The company will use the capital to scale its data-generation platform, expand its automated wet lab in Paris and launch an AI Virtual Cell programme.
What Does Rivercell Actually Build?
Rivercell builds cell-response data infrastructure and uses it to train models that predict how human cells react to drugs and genetic changes. Its automated wet lab runs perturbation experiments, exposing cells to compounds, genetic edits or environmental shifts. It records the results with live-cell imaging and gene-expression measurements, so each experiment yields time-resolved, multimodal single-cell data.
That data feeds a virtual cell model meant to simulate interventions in silico. The company's stated goal is to show that such predictions cut expensive bench work and produce hypotheses that later hold up in experiments. Rivercell has not published benchmark results, so that claim remains untested in public.
Who Is Behind the Company?
Yann Fleureau is co-founder and CEO. He previously co-founded Cardiologs, a cardiology AI company. Eric Durand serves as chief scientific officer, a role he took up in 2026. Durand worked as an oncology data scientist at Novartis, later served as chief data science officer at Owkin and co-founded Bioptimus.
The scientific advisory board includes Charlotte Bunne, an assistant professor at EPFL, and Fabian Theis of Helmholtz Munich. Both work on machine learning for single-cell biology. Fleureau has framed the opportunity bluntly: a world model of the cell is the largest opportunity in medicine, and the missing piece is data.
Why Does a Data-First Approach Matter?
A data-first approach matters because most published single-cell datasets are observational, not interventional. They show what cells look like, not how they respond when pushed. Models trained on that material can describe biology but struggle to predict the effect of a new drug or knockout.
Rivercell's wager is that generating its own controlled perturbation data, at scale and over time, produces a cleaner training set than anything assembled from public sources. The approach is expensive and slow to build, which explains the size of a seed round that would be large in most European biotech markets. It also ties the company's progress to lab throughput as much as to model architecture.
How Does This Fit European Biotech AI Funding?
The round fits a pattern of French and European investors backing AI-biology companies led by founders with prior exits or senior industry roles. Fleureau's Cardiologs history and Durand's Owkin and Bioptimus background make Rivercell a team drawn from the same Paris AI-health network. The participation of Bpifrance Digital Venture, the French state investment bank's venture arm, adds public money to a round led by a private fund.
The seed label deserves some skepticism. A €22M first round is closer to a Series A in size, which implies investors are paying for a lab build-out and a founding team, not for a product. With no disclosed valuation, revenue or pharma partner, the round says little yet about demand.
What Are the Risks?
The principal risk is validation. Virtual cell efforts across the industry have drawn criticism that models often fail to beat simple baselines when predicting responses to unseen perturbations. Rivercell will need to show out-of-sample accuracy that pharma teams trust before it can sell predictions or partner on programmes.
A second risk is business model clarity. The company could license data, sell model access or run discovery programmes with partners, and it has not said which it will pursue. Its scientific collaboration with EPFL gives it academic credibility, but no commercial partnerships have been announced.
Outlook
Rivercell enters the field with a seasoned team, a French and German-linked scientific board and enough capital to run its Paris lab for the next stage of data generation. The next milestones to watch are published validation of the virtual cell against held-out perturbations, the first pharma partnerships and any follow-on round that reveals a valuation. Until then, the €22M seed is a bet on data generation as the bottleneck in AI drug discovery.
