The war memorial on Marina Beach, Madras, against a heavy sky.
Est. 2026 · Madras, India

AI-native software, idea-smithed, hand-crafted.

We build intelligent software the way good things have always been built — by hand, in a small workshop. Veteran engineers with a decade or more behind them, and the people they are teaching, at the same bench. No account layers, no assembly line: the people who scope your work are the people who write it.

About
the madras software company
A workshop, not a factory. We take on a handful of engagements a year and stay with them long enough to be useful.

The Madras Software Company was founded on a simple, slightly old-fashioned idea: that software is a craft — and that crafts are learned at a bench, standing next to someone who has already done it.

Most software today is assembled — spun up fast, handed between teams, abandoned before anyone understands it. We work the other way around. Our teams are deliberately mixed: engineers with ten, fifteen, twenty years behind them, working alongside people at the very start of their careers whom they teach, review and vouch for. Nothing leaves here that a veteran hasn't put their name to. When something breaks at 3am, one of us knows why.

The work itself is AI-native — products where models do real work rather than decoration. But the method is unfashionably human: fewer projects, longer commitments, and code we are willing to sign. We build from Madras, a city that has been quietly engineering the world's systems for four decades, for clients anywhere who want what we want — software that still works, and still makes sense, five years from now.

What we do

Four kinds of work, done properly.

01

AI-native applications

Not a chatbot bolted onto a form. We design systems where models do real work inside the product — retrieval, extraction, agents, evaluation — and where the failure modes are understood before launch, not after.

02

Product engineering

Zero to launched. We take a product from a rough idea to something real people use — architecture, build, ship, and the unglamorous work of keeping it steady afterwards.

03

Platform & cloud

The layer nobody sees until it fails. Infrastructure, CI/CD, observability, and cost discipline — built so a small team can operate a serious system without heroics.

04

Rescue & rewrite

Inherited a codebase nobody can safely change? We map it, stabilise it, and modernise it in increments — without the eighteen-month rewrite that never lands.

How we work

Three commitments we do not negotiate.

Veterans lead, always

Every engagement is led by engineers with ten-plus years of shipping behind them. Newer engineers build beside them, never instead of them — every line reviewed, nothing unsupervised.

You own everything

Code, infrastructure, documentation, decisions. We write for the team that comes after us — including the one that isn't us.

Honest about scope

We would rather turn down work than pretend a timeline is real. If something is a bad idea, you will hear it from us early.

Careers

We hire at both ends.

Veterans with a decade or more who want to build and to teach — and people at the very beginning who want to be taught properly. Both, deliberately, on the same team.

Every apprentice is paired with a veteran from week one: real work, real review, no six months of shadowing. Being early here means shaping how the studio works rather than inheriting it. Madras, hybrid, genuine ownership. If nothing below fits but you think you belong here, write to us anyway — we read every note.

Write to us
Principal EngineerMadras · 12+ yrs
Senior AI EngineerMadras · 8+ yrs
Platform EngineerMadras · 5+ yrs
Apprentice EngineerMadras · Freshers
Product DesignerMadras · Contract

Open applications welcome year-round.

Contact

Tell us what you're building.

A paragraph is enough to start. We reply to everything within two working days — and if we're not the right studio for it, we'll say so and point you somewhere better.

hello@themadras.co