Tools that make everyday systems — language, courts, and markets — work for the people they actually affect.
What we build
d̪ɐni
ದನಿ · real-time call translation
A real-time voice translation app. On a live call each person speaks their own language and hears the other's, in both directions, as the call happens. You choose who you can call from inside the app — your phone's address book is never read.
An automated pipeline that turns the court's daily consolidated cause list into a public, plain-language dashboard — reconciling the matters that were scheduled against those actually heard. Every number is computed from the parsed data; the language model only writes prose over figures already calculated.
An analytics dashboard for the NSE Nifty 50 — open interest, Black-Scholes option pricing, implied volatility, institutional flows and daily market snapshots, with an agent that reads the day and writes it up.
The facts come from parsing and arithmetic. Language models add explanation on top — never the substance. If the model is unavailable, the numbers still stand.
Public sources only
We read what is already published, and we read it without touching the systems that publish it. No scraping of live infrastructure, no private data.
Cheap enough to keep running
Built to run at near-zero cost on static delivery and scheduled jobs, so nothing depends on a funding round to stay online.
One source of truth
Web and mobile read the same generated output, so the two surfaces can never quietly disagree with each other.