Why it mattered
Most pathology AI demos assume cloud comfort. Enso Atlas was deliberately built around a local-first premise: some environments need privacy, control, and interpretable evidence more than they need abstraction.
Pathology AI
Enso Atlas
A fully local pathology AI system for treatment-response prediction, evidence retrieval, and reporting on whole-slide images.
Designed to keep high-sensitivity clinical image workflows on-prem while still offering retrieval, explanation, and reporting.
What I built
Model orchestration across slide encoders and LLM components, a reporting surface for clinicians, and a retrieval layer that lets the system point back to relevant evidence instead of acting like a black box.
What I learned
Clinical credibility is not just an accuracy problem. It is a systems problem spanning latency, auditability, deployment assumptions, and how outputs are framed.
More work