CRASH Lab
Centre for Responsible Autonomous Systems in Healthcare · Koita Centre for Digital Health, Ashoka University · research group, founded May 2025 · crashlab.in
CRASH Lab is a clinician-led research group that builds and stress-tests autonomous and agentic AI systems for radiology and healthcare, with a focus on where these systems fail once they leave a benchmark and meet real clinical data — and on the evaluation and governance needed before they're trusted with a clinical decision. Our flagship programme, Radiology's Last Exam (RadLE), benchmarks frontier multimodal AI against board-certified radiologists on diagnostic accuracy, calibrated confidence, and readiness for autonomous use. The lab's other current focus areas are fine-grained diagnosis under label scarcity, agentic workflows for radiology reporting, and multimodal foundation models for image interpretation; see the Research page for the underlying work.
Lead
Suvrankar Datta, Founder and Group Lead. Simons Ashoka Early Career Fellow, Koita Centre for Digital Health, Ashoka University. About →
Roster
50 active members, including a 20-member core team (15 full-time), spanning roughly 30 Indian institutions. Named members and roles will be listed here as the lab's public roster page is built out.
Open positions
CRASH Lab is looking for research collaborators and students interested in agentic AI, multimodal foundation models, and responsible deployment of AI in radiology. If that's you, get in touch below.