Projects are grouped by how far along they are. Live means you can use it today. In research means we are still testing the idea and the numbers.
India-first PII masking and restore for LLMs. Catches Aadhaar, PAN, GSTIN, UPI and 18 more identifiers with checksums, and puts the real values back in the reply, even mid-stream.
An open benchmark for detecting Indian personal data: 2,500 documents across structured IDs, Hinglish, Hindi, Marathi, Tamil and Telugu, real bank codes, obfuscation, and look-alike numbers that should be left alone.
Guardrails for AI agents. Deterministic policy on tool calls: allow-lists, argument checks, limits on where data can be sent, and human approval for risky actions, with signed evidence of every decision.
A small model that finds Indian names and addresses in Indian scripts and code-mixed text. On IndicPII-Bench every tool we tested, including Saaf, scores near zero here today.
We pick problems where Indian data, languages or rules make the global tools fall short.