THC Community connect | Introduction to Clinical AI Evaluation & Rubric Design
THC AMA | 28th Feb | 11:00 AM – 12:30 PM
Theme: Introduction to Clinical AI Evaluation & Rubric Design
By: Garvita Baldua
Learn the basics of evaluation, how to define “good,” and how to catch silent failures before deployment.
Garvita spent years building AI dermatology systems at CureSkin, working closely with clinicians and real patient journeys. At scale, one thing becomes clear:
Models don’t fail loudly.
They fail quietly.
On specific skin types.
Under certain lighting conditions.
On edge-case severities.
On populations you didn’t test enough.
That invisible gap between demo performance and real-world reliability is exactly why she’s building Supertubos.ai — a clinical evaluation and data layer for healthcare AI focused on:
Clinician-verified ground truth
Failure-mode discovery
Slice-based evaluation
Designing models that are safe and dependable in real clinical settings
This session is not theory-heavy AI talk.
It’s a practical introduction to:
✔️ How clinical evaluation actually works
✔️ How to define what “good” means in healthcare AI
✔️ What to measure beyond accuracy
✔️ How to design an evaluation plan that clinicians and stakeholders can trust
If you’re building, deploying, investing in, or buying healthcare AI — this is a foundational skill.
The HealthTech Collective
THC is India’s largest community of >5,000 professionals for healthtech founders, operators, and innovators supporting early-stage healthcare innovation in India
Sign up here to join the community: https://nas.io/the-healthtech-collective-2