Senior Machine Learning Scientist - AI Virtual Cells Rivercell
Data generation infrastructure for AI Virtual Cell models in drug discovery.
CDI
Paris
Télétravail fréquent
Salaire : non spécifié
Les compétences que l’on cherche
Aptitudes à motiver les autres
Communication
Foundation
Pytorch
Votre nouvelle entreprise
Intelligence artificielle / Machine Learning
Pharmaceutique / Biotechnologique
Rivercell’s mission is to change how medicines are discovered by teaching AI to understand the human cell. Developing a single drug still takes years with an 90% failure rate in human. Predicting how cells respond to treatment before going to the lab would make it faster, and open up possibilities that are out of reach today.
That’s the promise of AI Virtual Cells: world models that learn the underlying rules of how cells respond to drugs and genetic changes, much as AlphaFold learned to predict protein structure. What holds them back is data. The vast, diverse, multimodal and time-resolved perturbational data they need doesn’t exist at scale, so we are building the infrastructure that generates it.
Our platform applies chemical and genetic perturbations to large cell populations and reads out their molecular and morphological responses at single-cell resolution, across millions of perturbation and context combinations. We then use that data to train our own world model of the cell.
Doing this takes a truly multidisciplinary team, where experts in thin films, microfluidics and instrumentation work side by side with experts in machine learning, computational biology and cell biology.
Vos missions au quotidien
The role
You will join Rivercell's team as a Senior Machine Learning Scientist, contributing to our work on AI virtual cells. You will develop and evaluate models, contribute to the technical roadmap, and work with colleagues on research and implementation.
The challenge
We believe the AI virtual cell, a world model that can predict how cells will respond before an experiment runs, is the biology challenge of our times. Rivercell's cell-on-a-chip technology offers the opportunity to develop models alongside the platform generating their training data. You will work with our biology and engineering teams to explore which experiments, data, and architectures make those predictions more reliable. Our goal is to make biological research faster and more predictive, with models that help researchers decide what to test next.
Implement, train, and test AI virtual cell models.
Contribute to the technical roadmap and research priorities through literature reviews and experimental findings.
Identify and harmonize useful cellular datasets with the biology team
Develop benchmarks and evaluation tools relevant to industrial users.
Maintain reproducible training code and analyses for the team and external collaborators.
Debug data pipelines and improve model performance and compute efficiency with engineering colleagues.
Contribute to preprints / publications, conference publications, and public benchmark releases.
Present findings and help respond to questions from reviewers and collaborators.
Support technical material for partners and fundraising.
Opportunity to work on a cutting-edge project in techbio
Opportunity to join an ambitious team and impactful startup among the first employees
Competitive salary and equity package
Exciting opportunities for personal and professional growth within the team
Ce que l’on cherche
You have a PhD in machine learning, computational biology, or an adjacent field, or equivalent industry research experience
You have hands-on experience training models on single-cell or perturbation data
You have at least one first-author publication in a relevant venue (Nature Methods / Cell Methods family, NeurIPS / ICLR / ICML main track or workshop, or a preprint with demonstrable adoption in the field)
You have strong PyTorch fluency, including distributed training on multi-GPU or multi-node setups
You are comfortable with the single-cell data engineering ecosystem and its real-world messiness
You have strong engineering practices: versioned data, pinned dependencies, reproducible analyses, documented decisions
You have strong scientific communication skills, written and spoken: write papers, present results to AI-lab and pharma audiences, and translate technical claims into investor-facing material
You have a make-it-work-make-it-better mentality: comfortable shipping the first version end-to-end on a pre-seed timeline, choosing pragmatism over elegance, and returning to harden the parts that proved load-bearing
Please let us know in your application if you specifically have experience with one or several of the following:
Experience with transformer or diffusion architectures applied to cellular data
Foundation-model pretraining, fine-tuning, or evaluation at scale
Familiarity with cellular morphology and image-feature pipelines
Benchmark design experience, including a sense of the politics of running a public leaderboard
Working knowledge of the current AI virtual cell literature and an opinion about where it is going
Experience in working in a fast-paced start-up environment
Nos étapes de recrutement
Throughout the recruiting process, you will meet 4 to 6 team members of Rivercell.
Introductory call (up to 30mn, videocall)
Technical interview (up to 90mn)
Performance interview (up to 90mn, in person)
Culture interview (up to 60mn, in person)