Seon Min Kim
Industry experienceAI evaluationConfidential

Clinical decision-support AI evaluation

Set the success criteria a clinical decision-support AI had to meet before use in care, and learned from clinicians what blocks adoption. Under a confidentiality agreement.

When
June 2025 – March 2026
Context
UCSF Innovation Ventures
Team
Team
My role
AI health technology evaluation intern
Tools
Evaluation design, Clinician interviews

Confidential

This work is covered by a confidentiality agreement, so I can’t name the technology or share specific results here. I’m happy to talk about my approach in an interview, within what the agreement allows.

What I did

I defined the success criteria the technology had to meet before clinical use: accuracy against confirmed diagnoses, the largest acceptable performance gap between patient groups, and response time.

What clinicians told us

Interviews with clinicians showed that EHR integration and changes to clinical workflow were the main barriers to adoption, more than model accuracy.

Outcome

Development of the technology continued at a startup.