Senior Product Manager

Patronus AI; San Francisco (CA); On-site; USD 180,000 - 280,000; Full time

Job summary

Patronus AI in San Francisco is seeking a Senior Product Manager to own the platform, infrastructure, and internal tools that let a small team turn frontier-lab demand into delivered RL environments. You will define a versioned, MVP-focused spec process and align with the OKR roadmap for engineering reliance on stable PRDs.

You will work with environment engineers, QA, and GTM to translate learnings into a ranked backlog, measure usage to guide decisions, and push for agentic workflows.

Qualifications

• 8+ years as a product manager with platform/infrastructure focus.
• Experience PMing products for engineers and internal teams.
• Track record of PRDs engineers actually built from.
• Strong technical fluency and ability to analyze usage data.
• Evidence of ruthless prioritization and clear written/ verbal communication.

Responsibilities

• Own the roadmap for the internal platform and tools end-to-end — environment platform, agentic build and QA tooling, expert onboarding, delivery tracking, and the spec pipeline.
• Write PRDs engineers trust and build from; versioned and frozen at kickoff with MVP scope.
• Collaborate with internal customers to identify pain points and translate into a ranked backlog.
• Make usage data the basis for decisions and kill features that don’t earn their keep.
• Close the loop between QA and spec; fix the pipeline rather than just the ticket.
• Apply agentic workflows for spec generation, review, and versioning; focus human hours on judgment calls.
• Ruthlessly scope; publish the cut list and iterate from evidence.
• Stay hands-on; run agents, prototype with AI, and build small evals to confirm features work.

Skills

Platform product management

Engineering collaboration

Data-driven decision making

Stakeholder communication

Education

BS/MS in CS or related

Job description

Patronus AI is a frontier lab developing simulation research and infrastructure to accelerate progress toward human-aligned AGI. We are on a mission to simulate all of the world’s intelligence.

We are the team behind some of the earliest and most influential research in AI evaluation like FinanceBench, Lynx, SimpleSafetyTests, CopyrightCatcher, Humanity’s Last Exam, and more. We are formerly AI researchers and engineers from companies like Meta AI, Amazon AGI, and Google. Our customers include foundation model labs and Fortune 500 enterprises like Adobe. We are backed by top-tier investors like Lightspeed Venture Partners, Notable Capital, Stanford University, Noam Brown, Gokul Rajaram, and more.

Responsibilities

As a Senior Product Manager at Patronus AI, you will own the product behind our production line: the platform, infrastructure, and internal tools that let a small team turn frontier-lab demand into delivered RL environments. That surface includes our hosted environment platform, agentic tooling for building and QA-ing application clones, expert onboarding, delivery tracking, and the spec pipeline that feeds engineering.

We run on a simple operating rule: if an agent can do it, don't assign it to anyone else. Your users are as often AI agents as they are people, and the products you spec should default to agentic workflows with humans as the exception. PMing for agents as first-class users is most of what makes this role interesting.

This is not a backlog-administration role. We believe you should not manage a product you don't use. Today, too many of our PRDs go stale the moment they're written, and engineers route around them; the product function sits on the critical path for the company's OKR roadmap. Your job is to make specs the thing engineers reach for first: versioned, frozen at build kickoff, scoped to an MVP with an explicit cut list, and grounded in real usage rather than opinion.

Your work will help frontier labs stress-test and improve the next generation of AI agents, advancing progress toward safe, human-aligned general intelligence.

In this role, you will:

• Own the roadmap for the internal platform and tools end-to-end — the environment platform, agentic build and QA tooling, expert onboarding, delivery tracking, and the spec pipeline. Sequence it against the company OKR roadmap so product is never the thing engineering is waiting on.
• Write PRDs engineers trust and build from. Versioned and frozen at kickoff, explicit about what's in the MVP and what's deliberately cut, and honest about buildability constraints (e.g., which classes of apps our agentic build tooling can and cannot handle today).
• Work with, identify the pain points of, and brainstorm solutions with internal customers — environment engineers, SME ops, QA, and GTM. Then turn what you learn into a ranked backlog and tell them plainly what's not getting built and why.
• Make usage data the basis for decisions. Instrument the tools, watch what agents and people actually do in them, and kill features that don't earn their keep.
• Close the loop between QA and spec. When automated QA keeps surfacing the same class of missing feature, that's a spec failure — fix the pipeline, not just the ticket.
• Apply the "if an agent can do it" rule to your own function: build agentic workflows for spec generation, review, and versioning, and spend your human hours on the judgment calls only you can make.
• Ruthlessly scope. Ship the smallest thing that changes an internal customer's week, publish the cut list, and iterate from evidence.
• Stay hands-on. Run agents in the environments, poke at the tools, prototype with AI, and build small evals to confirm whether a feature actually worked.

Qualifications

"The number one qualification to succeed in this machine learning course is gumption" - John Lafferty, CS Professor at Yale

Above all, we look for a proactive mindset, willingness to learn, unlimited energy, and relentless optimism. You are a great fit if you have a background in the following:

• BS, MS, or equivalent experience in Computer Science, Engineering, or another technical / quantitative field, with 8+ years as a product manager.
• Experience PMing platform, infrastructure, or internal tools — products whose customers are engineers and internal teams rather than external end users. This is required, not preferred.
• A track record of PRDs that engineers actually built from — and of shipping when priorities moved after the spec was written, without letting quality or trust in the spec slip.
• Strong technical fluency, including daily use of AI tools, comfort reading or reviewing code, and the ability to pull and analyze your own usage data. You should be able to go deep enough into the work to have credibility with engineers.
• Evidence of ruthless prioritization: MVPs shipped, cut lists published, and features you killed with the reasoning to back it up.
• Clear written and verbal communication, including the ability to turn messy stakeholder asks into a spec with crisp requirements, owners, and dates.
• Strong eye for quality and detail, with a bias toward catching gaps, inconsistencies, and subtle failure modes before engineering does.

Nice to have:

• Experience with reinforcement learning, agent evaluation, RL environments, or human-data / SME-sourced data pipelines.
• Experience building products whose primary users are AI agents rather than humans.
• 0→1 experience at a fast-growing startup.

To support close collaboration, this role is based in our San Francisco headquarters and requires in-office attendance 5 days a week.