Utility intelligence, audited
The intelligence layer for India's power distribution — that shows its work.
A working library of grid-AI capabilities that runs on a DISCOM's own data, speaks the regulator's language, and shows its work. Every answer carries where it came from, how it was computed, and whether it can be reproduced.
live modelled illustrative on-device
The grid, as one navigable substrate — every record labelled for what it is.
Why this is different
Every number can be traced, reproduced, and defended.
01
Capabilities, not dashboards
Sutra ships decisions and next-actions — not more charts to interpret.
02
Speaks the regulator's language
Framed in India's own grid vocabulary — PFC ratings, RDSS, the India Energy Stack — audit-ready, never a black box.
03
Shows its work, or stays quiet
Every answer carries its provenance and a witness. When Sutra can't ground an answer, it says so — it never invents one.
04
Runs on your data, built for IES
Adapter-first and IES-native. Connect a feed and the use-cases light up on your own grid.
The product
One platform. Three families you can license.
Pulse
Revenue & loss intelligence
Finds where unbilled consumption concentrates and makes the case your vigilance team can act on. Runs on your data once connected.
explore Pulse → Atlas
Smart-grid intelligence
Network risk, transformer health, and demand — worst-first, with the reasoning attached.
explore Atlas → Proof
Compliance MRV & disclosure
Turns the disclosure calendar into evidence you can file, reviewed before it leaves the building.
explore Proof → Why this isn't an LLM wrapper
How do you defend this answer when the regulator calls?
Sutra is built for the regulator's question. Every answer drills back to the row that produced it. Voltages and rupees are computed by the same formulas the regulator uses to audit you. And when nothing can ground the answer — Sutra refuses.
01
Defensibility. The trail is the deliverable — and it exists before the conversation starts.
02
Mechanical correctness. A load-flow result is the same today, next quarter, and three years from now.
03
Cost at scale. Routine questions take the local path; the cloud fires only when the question warrants it.
witness · replayed
answer — where unbilled consumption concentrates, worst-first
└ method · AMI disaggregation, stated formula
└ inputs · your meter feed, your billing export
└ rows · every record that produced the flag
└ if ungrounded · refuses, with hints
reproduce ▸ same inputs, same answer. always.
For partners — AMISPs · integrators · OEMs
Your brand, our engine, your environment.
You've laid the meters and won the bid. Sutra is the intelligence layer inside it — recognising IS 16444 / RDSS meter data in its native format, activating theft, asset-health, and forecasting use-cases on the feed you already operate.
Careers
Build the layer a regulator can replay.
Ontology, witnesses, refusal — some of the hardest honest-AI engineering anywhere, aimed at a grid a fifth of humanity runs on.
open roles →
Investors
The audited layer, at the policy moment.
Built for the RDSS and India Energy Stack moment — with a moat measured in witnesses, not prompts.
the thesis →
How we engage
Start with a paid pilot. If the after isn't better than the before, we refund you.
One bounded zone, forward-deployed, an exit clause from day one. The license that follows is flat per DISCOM — it does not grow as you add meters.