Careers
Come build the unglamorous parts.
The interesting work in AI is not prompt writing. It is evaluation harnesses, retrieval quality, cost control and systems that hold up unattended. If that sounds appealing rather than tedious, talk to us.
How we hire
One technical conversation, one paid work sample, one team fit call. No unpaid take-homes and no whiteboard algorithm theatre.
We care about how you reason through an ambiguous problem, whether you can say "I do not know" comfortably, and whether you have ever owned something in production long enough to be woken up by it. Framework familiarity is trainable; that judgement largely is not.
We are Mumbai-based with flexible remote arrangements for senior engineers in India. Open roles are not always listed — if you are strong and interested, write anyway. Tell us something you built and what went wrong with it.
What the work is actually like
Mostly evaluation harnesses, retrieval quality, cost control and failure handling — not prompt writing.
The interesting problems in applied AI are not the model. They are: why does retrieval return a confident wrong passage on this class of query, what happens when the provider rate-limits us mid-run, how do we make a retried tool call safe, and how do we know next month whether accuracy moved. If that reads as tedious, this is the wrong team. If it reads as the actual job, we would like to talk.
You will own things end to end. Small senior teams mean nobody hands you a ticket detached from the reason it exists — you talk to the client, you scope it, you build it, and you are the one who hears about it if it breaks.
How the interview runs
One technical conversation, one paid work sample, one team fit call. No unpaid take-homes, no whiteboard algorithm theatre.
The technical conversation is about something you have built, in depth — what broke, what you would do differently, where you got lucky. The work sample is a scoped piece of real work, paid at your day rate, usually four to six hours. We pay because asking for free labour filters for people with spare time rather than people who are good.
We aim to give a decision within a week of the last conversation, and a reason either way. Being ghosted after three rounds is a miserable experience and we will not do it to you.
The honest downsides
We are small, so there is no dedicated QA, no platform team, and some weeks the work is unglamorous maintenance.
There is also client contact, which not everyone wants. If your preference is to be handed well-specified work and left alone, a larger product company will suit you better and we would rather say that now than three months in.
Based in Mumbai, with flexible remote arrangements for senior engineers in India. Open roles are not always listed — if you are strong and interested, write anyway. Tell us something you built and what went wrong with it.
Tell us what you have built
Tell us what you are trying to build. We will tell you honestly whether we are the right team for it, and what it would realistically take.