Skip to content
Back to projects

01 / Context

The problem

Clinical AI research needs strict boundaries, traceable evidence, and an explicit refusal path when the retrieved context is insufficient.

02 / Implementation

What I built

Built bounded document and data agents, citation checks, synthetic FHIR data flows, and a local-model fine-tuning path using public or synthetic sources.

Current status
In progress
Evidence path
1 public link

03 / Current outcome

What exists today

A research environment for testing grounded retrieval and human approval gates without presenting clinical advice.

Boundary

What this page does not claim

Synthetic/public data only; non-clinical; not deployed as a clinical product.

Continue exploring

Full archive