Staff Product Manager, Agentic & AI Platform

Founded in 2019 and part of Y Combinator's 2020 cohort, Photoroom is the leading visual solution for e-commerce. We've raised Series B funding and reached 300+ million users worldwide, processing over 5 billion images annually and serving both small businesses and major enterprises like Amazon, DoorDash, and Decathlon through our mobile app, web platform, and API. We're a remote-friendly team of 100+ passionate builders giving e-commerce businesses superpowers to create visuals that help them grow, making the hardest parts of selling online disappear. We focus on craft, innovation, and collaboration, creating exceptional impact for e-commerce businesses worldwide.

Summary We're looking for a Staff Product Manager to own the AI platform every Photoroom product is built on. Our mission is that any team ships best-in-class GenAI visuals without needing a backend engineer: the primitives, the gateway and the no-code tooling turn adding an AI capability into a configuration step rather than an engineering project. On top of that sits the agentic layer — multi-step visual workflows that generate, check, reason, route and publish on their own. Working closely with engineers, ML researchers, and the product teams who ship on your platform, you'll decide what gets built, why, and how it ships. Location: Paris, (Hybrid) About the role Own the platform every AI feature runs through. Expand and improve our AI gateway, including safeguards, evals, cost and observability, giving teams the ability to ship with little to no code and build skills that we can ship directly to users. Build the agentic layer and Photoroom’s MCP. Visual agents that run multi-step flows: generate, QA, route, retry, publish, all unattended and at catalog scale. Define the problem before the solution. Sharpen the problem, then partner with engineering and research to ship work that feels effortless — for the developers and PMs using the platform as much as for end users. Treat other teams as your customers. Every team's roadmap contains AI features gated on platform work. You'll prioritise across them, hold a strategy rather than a request queue, and say no with reasons. Turn AI capabilities into products people want. Work hand-in-hand with our researchers and engineers to convert new model capabilities into shipped features that move usage and revenue, and turn them into primitives other teams can reuse. Use data to decide, not to justify. Pull and interpret data yourself, such as adoption, latency, inference cost per successful generation, eval results. Form hypotheses, measure what shipped, and call what's next. You won't need to wait around for someone else to run the query.

You've shipped products users love. 10+ years as a PM, with multiple launches where you have solved a user’s problem and delivered business outcomes. You've built platform, infrastructure or developer-facing products. APIs, SDKs, internal platforms, ML infrastructure — products whose users are engineers. You've shipped something agentic or model-backed into production. You understand evals, guardrails, non-deterministic output, cost per call, and what breaks when a multi-step flow runs unattended a million times. AI has reshaped how you work. You use Claude or equivalent every day, for prototyping, testing with users, or automating the parts of the job that used to fill your calendar. The value you bring sits in strategy and direction, not in writing tickets. You have demonstrable product taste. You can show specific work where the details matter and explain the calls you made, including the ones you'd make differently now. On a platform, taste shows up in the interface you make other people build against. You create clarity from ambiguity. Given a fuzzy problem, you take action rather than waiting for direction: you structure it, decide what matters, and move. You're data-fluent. You pull, interpret, and challenge data yourself, and use it to form a point