Careers
Innopas builds AI systems for financial services, healthcare, energy and government, industries where a wrong answer costs something and nobody accepts a demo as the deliverable. If you want to turn a vague idea into a platform that runs in production for a demanding customer, this is that job.
Two tracks
Most companies blur them and everybody loses, researchers get pulled onto delivery dates, engineers get asked to defend methods they did not choose. We hire, brief and measure the two separately, and people move between them deliberately.
You take a real operating problem, write it as a research question, and test an approach against a stated baseline with PhD advisors reviewing the method. Success includes writing up a negative result and stopping: that is a normal outcome here, and it is measured as one.
You take what survived and make it resilient, observable and secure enough to carry real load in a regulated environment. Judged on what still runs six months later, not on what demoed well.
Both tracks touch the same platform. Research ships into lalla.ai; engineering builds products on it. Whichever track you join, your work ends up in a product with a name. Not in a deck that closes a phase.
Early career
Our internships are launchpads, not placeholders. If we would not trust the work to reach a customer, we would not ask you to spend a summer on it.
Work that ships, with your name on it in the commit history. Shadowing is not a programme, it is a waiting room.
From people actively building in AI, cloud, data and security, not from a training function that left the tools behind years ago.
Our CEO makes time for early-career engineers because listening is part of leadership. Informal, candid, and about you: what you are learning and what is in your way.
Conversion when you demonstrate readiness, against expectations we state at the start rather than reveal at the end.
What you get
Stated plainly so you can hold us to it in your first month rather than discovering it in your first year.
Competitive compensation for your market. A culture that rewards learning and sharing through garages, tech talks and domain playbooks. Global exposure across blended teams in India, the US and elsewhere. Honest feedback and access to the people making decisions.
That you learn fast, ask uncomfortable questions, and treat the unglamorous ninety percent. The data model, the security review, the handover, and as the actual work. Most AI projects fail there, and it is where we need people who do not lose interest. If you want the longer version, read how the company got here and who you would be working with.
If nothing fits but you think you should be here, write to us anyway and say which track and which problem — that gets read.
Start here
Thirty minutes with the engineers who would do the work. No deck, no discovery invoice, a straight read on feasibility, sequence and what a first build would take, including when the answer is that you should not build it.
