Spiti

A reasoning model to help citizens identify schemes and benefits they are eligible for.

Heading bar

The challenge

A country may run over a thousand national and subnational welfare schemes — covering food, housing, education, health, and livelihood.

Each scheme comes with its own eligibility rules, varying by state, income, occupation, age, disability, and gender. Trying to find what you're eligible for is like trying to solve a really hard puzzle.

What if every person could discover exactly what they are eligible for - conversationally, in their own language, on their own phone?

A farmer in Vidarbha doesn't know she qualifies for PM-KISAN, Mahatma Jyotiba Phule Jan Arogya Yojana, and a state crop insurance scheme - all at the same time. A construction worker in Chennai doesn't know his children qualify for a scholarship he's never heard of. The schemes exist. The discovery doesn't.

Our Approach

We trained custom language models on the actual eligibility logic of every scheme we could find. Not summaries. The rules: who qualifies, under what conditions, with what documents, and what they receive.

The system asks a person five questions over WhatsApp — in their language — and returns exactly what they are eligible for. No app. No form. No literacy required.

22 Indian languages. Runs on a phone.

Spiti mobile demo
Spiti demo
Spiti sticker

The Work

Built for the rules, not the summaries.
The model doesn't retrieve information — it reasons through eligibility criteria the way a trained welfare officer would, checking conditions against what the person told it.

Local languages. WhatsApp. Voice.
Runs on a phone the person already owns, in either text or voice. The response comes back in the language they asked in.