Deriving the laws that govern biology
Numenos is a frontier AI lab building patient centric world models of human biology — and deploying them for therapeutic development
Physics and chemistry derived their laws through centuries of observation and mathematics. Biology defeated every attempt: too many variables, too much noise, too many interactions for any human mind to hold. Deriving its laws demands a new kind of AI —
architectures invented for biology from the ground up, operating at biology's own scale
Step One
Fingerprint a patient's unique biology
Every patient encodes information about how human biology works. We capture it — DNA, RNA, immune state, clinical history — and represent each patient as a biological system.
Step two
Scale across millions
One patient is a data point. Millions are a map. Our representations span populations, diseases, therapies, and time — and what looks like noise in isolation becomes structure at scale.
Step three
See biology as a system
Disease doesn't live in silos. We connect patient representations into one continuous model that learns across contexts — the way neurons bind separate sensations into a single experience. With the right architectures mechanisms surface and Biology reveals how it actually operates.
Many diseases. One biology.
The silos in drug development were drawn by org charts and anatomy, never by biology. Data is split by indication and organ, so every program starts from zero. Our architecture learns from patient biology as a whole, across every disease at once, so knowledge stops resetting and starts compounding. What emerges is the biology diseases share, and the context that decides when it matters.
Biologically informed Cross-trial
optimization, publication with Boehringer Ingelheim
Learning from one lung cancer trial
enabled optimization of a second, never- seen trial, turning a missed endpoint into a significant result while keeping 88% of patients and increasing statistical significance ~10x. Biology learned in one trial improved decisions in the next.
Patient-level latent space representations identify key target in
Atopic Dermatitis
In atopic dermatitis, Numenos’
individualized modeling elevated IL-13 from a marginal signal (#265 in
traditional analysis) to a top driver (#9), clarifying which patients are truly driven by IL-13 biology and enabling precise target and trial design.
Biologically informed Cross Indication
Prediction
Trained on kidney cancer patients,
Numenos’ model identified which lung cancer patients benefit from immunotherapy, turning a negative NSCLC trial positive with 35% fewer patients and showing that biology learned in one cancer transfers to another.
Harmonized, annotated multiomics + outcomes with governed provenance
- Annotating data at scale enables fast iteration on model building
- Adaptive regulatory framework for traceability, transparency, explainability of data
Synaptic patient representations that learn biology continuously across diseases
- Patient centric biological representations that optimize for clinical trial success to find unmet need biology from target inception
- Breaking biological silos across therapeutics areas and organizations
Decision systems that translate biological intelligence into clinical and commercial outcomes
- Identify coherent biological subpopulations
- Maximize commercial potential of every asset across disease areas
- Identify optimal indication, patient population and synergistic drug combinations
Built by Cross Disciplinary Team Spanning Frontier AI, Biology & Pharma
Team
Vitalay Fomin
CEO & Co-founder
Amit Weiss
Chief Technology Officer & Co-founder
Neil Pfister
Chief Medical Officer
board members
Vitalay Fomin
CEO & Co-founder
Amit Weiss
Chief Technology Officer & Co-founder
Yaron Samid
Board Member
Daniel Roditi
Board Member
advisory board
David Feltquate, MD, PhD
CMO
Mike Branson, PhD
SVP, Head Biometrics and Data Science at UCB
Ralf Halbach, PhD MBA
CEO, Head of Avastin lifecycle
Allan Shaw
CFO
Dvorit Samid, PhD
EVP, Medical Affairs