A continuous-time multimodal foundation model for human physiology. The inference layer, not the sensor layer.
Drug development is broken at the biology layer.
Nine out of ten drugs that enter clinical trials never reach patients. The most common reason is not commercial — it is biological. Toxicity that was invisible at the target level becomes catastrophic at the system level.
Current models test compounds against isolated targets. The human body does not work that way. The liver talks to the heart. The kidney talks to the liver. A compound that looks clean in isolation can trigger a cascade no single-organ model will ever predict.
Lumos is building the physics engine for the human body — a foundation model trained on continuous physiological time-series that sees the whole system, not just the target.
PhysioFM reads the body
as a continuous signal.
The inference layer for
human physiology.
“The problem is not that we lack data. The problem is that we have never had a model capable of reading it at the level of the whole organism. That is what we are building.”
Built by people who understand the problem.
Our team comes from Harvard, Princeton, Stanford, and UC Berkeley — with backgrounds spanning computational biology, machine learning, drug development, and patient-facing clinical care.
Coming Soon
We're working on sharing our research and insights. Check back soon for updates on our work in computational physiology.
We're building the team.
We are a small team working on a hard problem. If you have a background in ML systems, computational biology, pharmacokinetics, or clinical data — we want to hear from you.
Let's talk.
Whether you're an investor, a potential collaborator, a pharma partner, or someone who wants to work on this problem — reach out directly.