Senior Product Designer (AI Native)

Read this first

You have shipped product design that moved a real metric at scale. And you personally build with AI tools: working prototypes, not static frames, and at least one AI-native experience that reached real users.

If both are true, everything below will read like the role you have been waiting for. If not, stop here. We mean the filter literally: the first interview is you demoing something you built yourself, running live, on your machine. Not a case study. Not a walkthrough of frames. A working thing.

The context

Jupiter is a neobank across savings, payments, cards, loans, investments, and insurance, built on RBI-regulated entities with partner banks. We have made a company-level bet: AI is not a feature layer on the app, it is the next version of the app. A family of production agents already runs inside the business, and the next phase is customer-facing, starting with an agent that replaces our customer servicing stack.

But most of what users touch this year looks nothing like a chatbot. Onboarding. Card activation. Repayment flows. Payment success screens. The designers who win here carry both: classic consumer design craft where a two percent funnel lift is real money, and the ability to design for a model whose output is not deterministic. And here is the genuinely unsolved part: when an agent moves money, blocks a card, or explains a regulatory protection, the cost of a bad interaction is real financial harm. The design patterns for this do not exist yet. Confirmation without friction. Confidence without overclaiming. Disclosure that satisfies a regulator and still reads like a human wrote it. You will be creating these patterns, not applying them.

What you'll own

Core product surfaces. Depending on fit: onboarding, payments, cards, or lending. Funnels, friction, comprehension, trust. You own design metrics on these surfaces and you move them.

The AI interaction system. A coherent design language for every agentic surface at Jupiter: conversation, streaming, tool-use visibility, citations, confirmations, error and repair states, escalation to humans. One system, so a Jupiter agent feels like a Jupiter agent whether it is answering a KYC question or walking someone through a failed UPI transaction.

Trust and safety surfaces. The moments where an agent is about to act on money or data. Consent, confirmation, and audit patterns that let users move fast and still feel in control. This is where design earns its seat in a regulated business.

Your own leverage stack. This is where the 10X comes from. You scout, evaluate, and wire together the tools that change what one designer can produce: prototyping setups, agents for design audits and first-pass explorations, transcript review tooling, automated accessibility and consistency checks. We do not hand you a stack. We expect you to bring one, keep upgrading it as the tool landscape shifts month to month, and turn the best of it into the default for the whole design org. What you build for yourself often becomes what everyone uses.

How you'll work

Your designs ship first as working prototypes you build yourself, with a real model, real latency, real failure modes. We test the thing, not a click-through of the thing. On agentic surfaces, you write prompts and shape system behavior alongside product and engineering, because the model's behavior IS the interface. You read transcripts of real agent conversations the way a traditional designer reads session recordings. On classic surfaces, you live in the funnel data and treat every drop-off as a design question before it becomes a growth question.

And you keep judgment where judgment belongs. AI does your grunt work: explorations, audits, first drafts. It does not make your calls on hierarchy, trust, or what a moment should feel like. The leverage exists to buy you more time on exactly those calls.

What you won't do

Hand off static frames and move to the next ticket. Design happy paths and let engineering improvise the other ninety percent. Split the world into AI design and normal design and pick a side. Wait for a research team to tell you what users felt. Ask for a bigger design team before asking whether a tool you could build closes the gap.

What success looks like at six months

A metric on a core surface you own has moved, and you can trace exactly why. The AI interaction system exists, is documented, and every new agentic surface at Jupiter is built on it. At least three of your designs went to users as prototypes you built yourself before an engineer wrote production code. Your leverage stack is real and measurable: work that took the team days now takes hours because of tools you brought in or built, and at least two of them have been adopted by other designers. When someone at Jupiter asks how should the agent behave here, they ask you. When someone asks is there a tool for this, they also ask you.

How we'll interview you

1. The demo. Bring something AI-powered that you designed and built yourself. Running, not recorded. We will use it together and change things live.
2. The number. Walk us through one classic design project that moved a metric at scale. We will push on attribution and what you got wrong.
3. The stack. Walk us through your current tool stack: what you use, what you dropped, and why. We are listening for taste and velocity, whether you evaluate tools against real work or against hype.
4. The teardown. We give you a real transcript of an agent conversation that went wrong, and a funnel with a drop-off. You tell us what the design should have done in each. Same muscle, both worlds.
5. The build. A short working session where you prototype a confirmation flow for an agent action, with AI tools, in front of us. We care about how you think with the tools, not pixel polish under pressure.

Logistics

Bangalore, in person. Five to seven years of experience in consumer product design at scale, fintech a strong plus. Reports into the design org, works daily with AI product and engineering leadership. ₹40L to ₹70L cash, plus equity, higher for exceptional candidates.