A small opinionated
AI practice — for
industries that can't
afford generic.
We're 52 engineers, researchers and domain leads. Engineering centre of gravity in Howrah, India. Client teams in London and Wilmington, Delaware. We don't take projects we don't know how to ship.
Six rules we won't break.
Domain > model.
A frontier model is a commodity. The seven years your operators spent learning ICD-10 isn't. We start with the domain — the model comes second.
Outcomes > outputs.
Every engagement starts with the metric we're moving. If we can't write it on a sticky note, we don't take the project.
Boring > clever.
The unsexy data pipeline always wins over the elegant model architecture. We over-invest in evaluation, monitoring, lineage and the long tail of failure modes.
Pilots > decks.
We don't sell six-month strategy phases. Every two weeks there's a working artefact your stakeholders can poke at.
Your cloud, your keys.
We deploy inside your VPC, your tenant, your KMS. The model is our work; the data is yours. We do not pool client data, ever.
Honest about limits.
If we don't think AI is the right answer, we say so — even when you're paying us. We've recommended SaaS, hiring, or doing nothing more often than is good for the P&L.
Eight years.
Same thesis.
Founded in Howrah
Started as a two-person consultancy: one ML engineer, one ex-bank ops lead. First contract was a fraud-scoring model for a regional NBFC.
First international clients
UK and US offices opened. First healthcare and legal engagements. The "domain > model" thesis crystallized after losing a competitive pitch to a generic SaaS that did fine for 70% of the problem.
The pivot to vertical AI
Stopped doing generic ML work. Built six dedicated industry practices. Hired domain leads as full partners — not advisors.
ISO 27001 + 100th client
Hit 100 clients across 14 countries. Achieved ISO 27001 certification (monitored via Sprinto). Launched the managed-AI tier for production support.
Where we are now
52 people, six verticals, three offices. About 60% of revenue from clients we've worked with for two or more years.
Five percent of capacity,
every quarter, on the public good.
We run a standing SDG cohort: climate startups, public-health orgs, education non-profits. Same engineers, same standards, sliding-scale price. Recent ships include a carbon-emission calculator and a recycle-bin classification app for municipal pickup.