Careers at Sutra
Build the layer a regulator can replay.
Most AI teams ship answers. We ship answers that survive an audit — grounded in an ontology of India's grid, witnessed row by row, and honest enough to refuse. That is a harder, rarer engineering problem than another chat wrapper. It's why we exist.
The problem, honestly
Why this is some of the hardest applied-AI work going.
the substrate
One semantic model of the grid
Meters, feeders, transformers, tariff orders, ratings — one ontology every capability reads from. You'll shape how a nation's grid is described to machines.
the contract
Every answer carries a witness
Provenance isn't a feature we bolt on — it's the invariant everything is built against. Deterministic, replayable, defensible when a filing is contested.
the discipline
Teaching a system to refuse
The recurring verb here is refuse. Making an AI stay quiet when the evidence isn't there is harder than making it talk — and it's the whole product.
How we work
Forward-deployed, evidence-first, small on purpose.
You'll sit with the desk that signs. Engineers travel to the DISCOM. The vigilance officer's Monday is your spec.
Decisions are written down. Architecture decision records for everything that matters — you inherit the reasoning, not folklore.
Open-source substrate. The deployed stack is permissive open-source. We build in the open idiom, on our own terms.
Honesty over demo-magic. Live, modelled, illustrative — labelled on screen, even when a sale is at stake.
The grid a fifth of humanity runs on deserves software that shows its work.
Sutra.
The audited utility intelligence layer for India's power distribution.
livemodelledon-device
Products
PulseAtlasProofRegulatory BenchmarkingProcurement & Markets
Company
PlatformWhy SutraCareersInvestors
Engage
How we engageBook a pilotFor partnersContact