19 Nov
|
Jupiter
|
Bengaluru
About Jupiter
Jupiter is a money app for people who are building their lives in India. From one app you can open a savings account with our partner bank, pay on UPI, save towards goals in Pots, borrow, buy insurance, invest, and see where your money actually goes. Our Edge+ CSB Bank RuPay credit card works on UPI and earns Jewels, our rewards currency, on the brands people already shop with.
We are a product company at heart. We obsess over small details, we read what customers tell us, and we would rather do fewer things properly than many things badly.
Why this role exists
AI is already part of how Jupiter works. Our product team uses an in-house assistant that reads our Confluence, Figma, Amplitude, Databricks and support tickets to answer questions and draft PRDs, right inside Slack. Engineering has a bot that turns a Jira ticket into a draft pull request. Marketing has a creative studio that designs ads, critiques its own work against a checklist and tells us what every run cost.
Those started as tools for us. The next step is bigger: putting AI in front of customers in a product that handles their money, where a confident wrong answer is not a small bug. We are looking for a PM who wants to own that problem, with the judgement to know where AI helps and the discipline to prove it before it ships.
What you will do
- Find the problems worth solving with AI. Explaining a declined payment, making sense of a card bill, help that actually closes the ticket, a Money tab that answers questions in plain words. You will also say no to the ideas where a better screen would do the job.
- Own products end to end. Problem, PRD, evals, launch, and the numbers after launch. You stay with it until it works for customers, not until it is merged.
- Write the evals before the feature. Define what good looks like with real examples, build the test sets with engineering, and decide the score that is good enough to ship. When the model changes, you know within a day whether things got better or worse.
- Design for the moments AI gets it wrong. What the assistant says when it does not know, when it hands over to a person, and how a customer corrects it. These moments decide whether people trust us.
- Keep it safe and compliant. Work with compliance, risk and our partner banks so everything we build respects RBI guidelines, the DPDP Act and our partners' policies. Customer data stays where it belongs.
- Treat cost and speed as product decisions. Know what each conversation costs, pick the right model for each job, and cut what does not earn its keep.
- Raise the bar for the whole team. Help other PMs, designers and analysts use AI well in their own work, and keep improving the internal tools they already rely on.
Your first 90 days
- By day 30: you use Jupiter every day, have read a few hundred support conversations, have sat with ops and compliance, and know our internal AI tools by using them.
- By day 60: you have picked one customer problem, made the case for it with real data, and put a working prototype in front of colleagues.
- By day 90: you have an eval set, a launch plan with guardrails,
and a small rollout to real customers with one clear success metric.
You will do well here if you
- Have 4 to 8 years in product, with at least one AI or ML product shipped to real users. A demo does not count; something people used does.
- Can explain in plain words how language models fail (making things up, prompt injection, behaviour shifting between model versions) and how you would design around each.
- Are comfortable writing prompts, reading a model's outputs in bulk and spotting the pattern in what went wrong.
- Use data without being asked. You can write SQL, build a funnel in Amplitude, and tell a real change from noise.
- Write clearly. Your PRDs are short, specific and something an engineer can build from.
- Care about the person on the other side of the screen, especially when the subject is their money.
Positive to have
- Experience in fintech, payments, lending or another regulated space.
- You have built with the Claude or OpenAI APIs yourself, even on a side project.
- Hands-on time with retrieval (RAG), agents that use tools, or eval tooling.
- You can sketch a flow in Figma before a designer picks it up.
What you will work with
- Claude and other frontier models through our AI gateway, and MCP to connect agents to our systems.
- Amplitude, Databricks, Jira, Confluence, Figma and Sense UI, our design system.
- A React Native app, and engineers, designers and analysts who like to move quickly and measure what they ship.
How we work
- Small teams own outcomes, not tickets.
- Decisions are written down with the reasoning, so anyone can pick up the thread later.
- We are honest about results. If a launch did not work, we say so and fix it.
📌 AI Product Manager (Bengaluru)
🏢 Jupiter
📍 Bengaluru