The Sandhi FieldFine threads from two directions meet at a single junction.

SANDHI Research Lab

Where ideas meet, discovery begins.

An independent research lab exploring artificial intelligence and computational science. We investigate difficult problems across vision, language, multimodal AI, and AI for science.

सन्धिsandhi: the place where two things join.

At the boundary, both sides change.

In Sanskrit grammar, sandhi describes what happens where two words meet. Sounds shift, merge, and settle into something that reads as one. We think research works the same way. The most useful ideas rarely come from a single field. They form at the junctions, where methods, disciplines, and people meet and are changed by the meeting.

Read about SANDHI

Our research philosophy

  1. Connect

    We bring together ideas, methods, and people from different fields.

  2. Investigate

    We pursue questions where existing methods fall short.

  3. Understand

    We test rigorously, so that results explain rather than merely impress.

  4. Contribute

    We publish, release code and data, and add knowledge others can build on.

  1. Perception

    How machines see

    Computer Vision

  2. Language

    How machines read, reason, and speak

    Language Models & NLP

  3. Junction

    Where modalities meet

    Vision-Language Models, Multimodal AI

  4. Understanding

    What is learned, and why it holds

    Representation Learning, Causal Inference

  5. Discovery

    Computation in service of science

    AI for Science, Computational Biology

Featured projects

All projects

Our first projects will be published here soon.

Recent publications

All publications

Our first paper will appear here soon.

From the lab

All news

News from the lab will appear here.

Research with us.

We look for people who want to investigate difficult problems carefully and contribute knowledge that lasts.

Join SANDHI