Forward Deployed Engineer, Zentis AI (by Techvantage.ai)

Forward Deployed Engineer - please read the JD fully before applying

Agentic AI for regulated BFSI. Five markets. The person who takes it into production.

Trivandrum, India, hybrid, with travel to client sites across our markets

About the job

Take the AI that audits a bank's books, underwrites a policy, or reads a claim file, and make it work inside the client's own walls. In production, under regulatory scrutiny, with real money on the line. Not demos. Not pilots that die in a slide deck.

The Forward Deployed Engineer closes the distance between a demo and a system a top tier institution will actually run. You sit with their audit, actuarial, finance, and credit teams. You learn how they really work, not how a slide says they do. You shape our platform to fit, you get it live, and you carry what you learn back into the product. This is build in India, for the world work, deployed into leading banks, insurers, and NBFCs across India, the GCC, Southeast Asia, the UK, and the US.

This is not a services role and it is not a sales engineering role. You build. You ship. You own the outcome at the client, and you own making the platform better because of it.

What you will do

• Take named agents inside our SPAR framework the last mile into production, inside the client's environment, not a sandbox.
• Embed with client teams and shape SPAR agents to a specific domain, whether that is bank audit, claims intelligence, insurance forecasting, or retail credit.
• Build and calibrate evals that catch the failure modes a risk officer cares about, not the ones a benchmark measures.
• Architect retrieval, memory, and tool use so a high stakes decision is auditable after the fact, not just correct in the moment.
• Turn one client customization into a product decision. Decide with the team whether it becomes a configurable capability or stays a considered one off. You do not let client work quietly fork the product.
• Be the fastest feedback loop we have. What you see in the field sets the roadmap and feeds the geo variant matrix.
• Bring creative inputs and dream about the product with us. The best agent designs here started as someone's what if we tried. We want people who brainstorm out loud and argue for the better idea.

What you bring

• 6+ years in ML and AI, with 3+ taking models from prototype to production. For this seat, production means inside someone else's regulated environment, with their data and their auditors, not your own.
• Real range across the discipline: statistics, classical ML, deep learning, NLP, and now LLMs and agentic systems. You understand why things work, not just which library to import.
• Strong Python, with real production experience shipping LLMs and agentic systems, and enough front end range to build the surface a client sees.
• Process discipline. In regulated AI, the discipline is the product: repeatable, documented, defensible.
• Clear writing, and the range to explain an agent's decision to an engineer, a risk officer, and a CTO alike. In this role you are in the room with all three.
• The willingness to travel and to sit on site. Forward deployed means forward. The learning happens in the client's building, not over a call.
• The drive to keep learning at the pace this field moves. What is frontier today is table stakes in a year, and you stay ahead of that on your own.
• Judgment about what to productize and what to refuse. This is the part that separates a Forward Deployed Engineer from an expensive consultant.
• Bonus: fine tuning, eval frameworks, DMN or YAML rule engines, BFSI domain.
• Bachelor's or Master's in Engineering, Computer Science, IT, or a related field, or a body of work that makes the degree irrelevant.

On AI coding tools: we use Claude and others to review and accelerate work we fully understand, never to ship code we cannot defend. We build proprietary IP. Judicious, not reckless.

A quick word on fit

• This matters, because it saves us both time.
• This is not an easy role, and we are not looking for people who want one. It rewards people who will toil and go the extra mile to get real AI into production, and who are at their best when the stakes are real and the client is watching.
• If you love hard problems, if you get restless when you are not learning, if the problem itself is the reason you would take the job, you will thrive here.
• It is a harder fit if you are looking for an easy seat, or a role where the work stops at the demo, or a brand for your CV. None of that is wrong. It is just not what this seat is for.
• In return you get experience you cannot buy. Production AI running inside regulated institutions. Cross domain exposure across banking, insurance, and credit in five markets. And the kind of founder proximity where your judgment actually shapes what the product becomes. We back that kind of person with real equity and real ownership.
• You will know very quickly which one you are. So will we.

Why Zentis

Agentic AI for the regulated enterprise, incubated by Techvantage.ai, selling into top tier institutions where most AI never clears compliance and ours is built to. ESOPs from day one. Measurable ROI in production, with customers who become references. Founder proximity, frontier problems, and a seat in the room when we decide what to build next. We move fast, we argue hard about the right design, and we hold a bar most places only talk about. If that is how you like to work, you will fit.

How we work

• We measure what you ship, not the hours you log. That is real autonomy for people who move, and little patience for those who wait to be told. We hire slowly and choose carefully. We pay well and fairly for the right person, because we are not chasing numbers. We are not interested in inflated titles or packages that no one can justify. We care about what you can actually do.
• This is demanding work, in a hard domain, for people who take accountability and stay loyal to the goal even when it is difficult. It is not a pit stop. If you are optimising for the highest possible package rather than the work, this probably is not your seat, and that is okay. We would both rather find that out now.
• Hybrid: three days together in the office, two flexible, plus travel to client sites when a deployment needs you there.

Apply

Email your updated CV to [email protected]. Tell us, in a line or two, about the hardest thing you took all the way into production, what nearly broke on the way, and how you got it live anyway.