Clinical trial qualifier

slide 1
slide 2
slide 3
slide 4
slide 5
slide 6

Inspiration

The volume and pace of new cancer trials outstrips what clinicians can manually track. The gap between eligible patients and available studies is avoidable.

What it does

Automates PHI redaction, analyzes patient records, retrieves real-time trial data, and programmatically checks inclusion and exclusion criteria.

How we built it

Combined PHI detection, an AI reasoning agent, and an MCP-linked ClinicalTrials.gov pipeline inside a FastAPI service with structured evaluation logic.

Challenges we ran into

Unstructured EHR text, variable trial criteria formats, slow multi-trial evaluation, and maintaining strict privacy handling. The MCP was difficult to set up as well

Accomplishments that we're proud of

Reliable automated redaction, accurate eligibility reasoning, and real-time sourcing from ClinicalTrials.gov with consistent structured outputs.

What we learned

Clinical criteria are more ambiguous than expected, and rigid parsing fails; AI reasoning layered over structured checks is required for usable results.