Orchestration, not a single vendor
The product is not wired to one LLM. Clinical-language models, biomedical literature models, embeddings, genomics and molecular tools are separate adapters. Public APIs are not assumed to be approved for health data.
Models and research systems we may evaluate
Candidate technologies evaluated for specific research tasks. Inclusion here does not mean a model is approved for clinical decision-making or authorized to receive identifiable patient data.
Google MedGemma — health/medical language and multimodal research workflows. Not claimed as a diagnostician.
Microsoft BioGPT — biomedical literature NLP. A research component, not clinical authority.
NVIDIA BioNeMo ecosystem — biology and drug-discovery research tools rather than patient-facing diagnosis.
Approved general-purpose models and specialized genomics, protein and molecular models may be evaluated under the same governance.
Training
Patient data must not automatically train any model. Separate research-data authorizations are required. Joining does not require contributing health information to AI training.