Founder and research lead
SANDHI Research Lab
Where ideas meet, discovery begins.
An independent research lab exploring artificial intelligence and computational science — difficult problems across vision, language, multimodal AI and AI for science.
सन्धि — the place where two things join
In Sanskrit grammar, sandhi is what happens where two words meet: sounds shift and settle into something that reads as one. The lab is built on the same bet about research — that the useful ideas rarely come from a single field, and form instead at the junctions, where methods, disciplines and people meet and are changed by the meeting.
Five themes
- Perception
- Computer vision
- Language
- Language models and NLP
- Junction
- Vision-language models, multimodal AI
- Understanding
- Representation learning, causal inference
- Discovery
- AI for science, computational biology
The lab's platform
I built the lab's site as a single application: the public face — the research map, projects with public progress, publications, notes and applications to join — together with a members' workspace for tasks, meetings, experiment logs and drafts sent for review, and the administration behind both. Invitation-only accounts with two-factor sign-in for everyone, every administrative change audited in the same transaction as the change itself, 406 unit tests and more than 140 end-to-end tests, and fourteen days from the first commit to production.
Everything about the lab — its research, publications, people and how to join — lives on its own site.