Building open-source tools for reliable, transparent and auditable medical AI.
My current open-source project is Clinical AI Reliability (CAR), an experimental framework for creating structured reliability records around medical and health AI outputs.
CAR focuses on:
- claim-level evidence tracking
- evidence freshness
- uncertainty reporting
- counterfactual consistency testing
- privacy-minimizing provenance
- human-review triggers
- reproducible AI evaluation
- machine-readable Reliability Envelopes
CAR is research infrastructure. It does not diagnose patients, replace clinical judgment, or certify AI systems as medically safe.
CAR v0.1.0 is available as an early public research pre-release.
Medical AI • Health AI • AI Reliability • AI Evaluation • Provenance • Counterfactual Testing • Reproducibility • Open Science
Feedback, research ideas, benchmark proposals and contributions are welcome through the project's GitHub Discussions.
Sabrina Chowdhury Tithi