Live
Read now

Lumos PBC

Making physiology computable.

health AIdigital healthmedtechfoundation models in physiology
StageSeries B
Raised$48M
Founded2026
Employees6
LocationPrinceton, NJ

About

Lumos Health builds AI that predicts how drugs affect multiple organ systems simultaneously — cardiac, renal, hepatic, respiratory, metabolic — from continuous physiological data. Every Phase I clinical trial monitors safety with disconnected, single-timepoint measurements analyzed weeks later. Lumos processes the same data your trial already collects — continuous ECG waveforms, vital signs, lab results, drug administration records — and predicts multi-organ drug effects in real time. Cross-organ interactions that single-endpoint monitoring misses become visible hours to days earlier. Our foundation model was trained on 42,000+ ICU patients and 200+ drugs. It learned how human physiology responds to pharmacological intervention by ingesting continuous waveforms fused with clinical data. In validation, the model discovered known drug-organ interactions — nephrotoxicity, respiratory depression, cardiac perturbation — in a fully unsupervised setting. No labels. No prior pharmacological knowledge. Just physiological signal. Built for the drugs that need it most: novel mechanisms with no historical safety database. RNA therapeutics. Gene therapies. Combination regimens. If there is no reference compound to benchmark against, Lumos provides the computational safety baseline. Founded by a Princeton mathematician, backed by pharma and healthcare investors, and advised by clinicians from Penn Medicine, Princeton, and Harvard. We are accepting early access partners for Phase I safety monitoring. Contact: zach@projectlumos.org

Product

Value proposition
Lumos is building a foundation model of human physiology—a continuous-time, multimodal model that learns how the body actually works from raw physiological signals, enabling continuous physiological data understanding for medicine and drug development.

Founders

  • Rick Hawkins
  • Jon Saxe

Categories

foundational AIphysiology modelingmedical AImultimodal time-series

Key features

continuous-time multimodal physiological modelingraw signal learningmodular encoders for each sensing modalityshared backbone for unified patient state representationplug-in new sensors without retraining

Technologies

foundation modelsmultimodal time-seriescontinuous-time modelingencoder architecturesbackbone representations

Partnerships

Partner with us on continuous physiological data.

Investors

New Enterprise AssociatesSanté VenturesDeerfield

News

Browse similar startups

Related startups

Profile updated Jul 29, 2026. Sources: projectlumos.org

ClinAdsby ClinsightBuilt for clinics

Your clinic’s marketing team, powered by AI.

On-brand ads, SEO, and social campaigns in one place.

Generate your first clinic ad for free
Clincyby ClinsightBuilt for practices

Get more revenue from the patients you already have.

Voice and SMS that follow up until lapsed patients are back on the schedule.

Book a Clincy call
DDS RecruitPowered by ClinsightDental recruiting

Your practice’s fractional recruiting team.

Dental-only search, run end to end, with placements guaranteed for a year.

Find your next hire