Lead Product Manager, Agent Runtime & Evals

junipersquare; United States; Remote; USD 150,000 - 210,000; Full time

Job summary

Juniper Square is seeking a Lead Product Manager to own the agent runtime and evaluation layer of Foundry, the platform for AI agents, apps, and code. It is built for non-engineers who can describe needs and receive governed automation with identity, access, and audit trails.

You will report to the Group Product Manager for Foundry and collaborate with a dedicated engineering team. About 75% of your time focuses on the agent runtime—versions, deployments, evaluations, and handling model

Responsibilities

• Build the agent runtime for non-engineers: define the vision, strategy, and roadmap for a platform where business users take an agent, app, or script from idea to governed production without an engineering cycle.
• Own evaluations and model governance: define testing, golden datasets, labeling and regression gates; decide when models are eligible for agents and when human oversight is needed.
• Address the 'who evaluates the evaluator' challenge: ensure checks for agents that grade other agents, partnering with engineering on robust patterns.

Job description

About Juniper Square

Private markets are one of the largest, most complex, and most underserved corners of global finance. Our mission at Juniper Square is to unlock their full potential. We're the Operations Partner trusted by 2,300+ GPs, unifying technology, data, and fund administration services into a single platform that helps GPs move faster, make better decisions, and scale with precision. With $300B+ under administration and 700,000+ LPs on platform, we've built the scale to match our ambition. And with JunieAI, our purpose-built AI platform, we're reimagining how private markets operate, embedding intelligence across every workflow. Founder-led since 2014, backed by $350M+ in funding, and now 1,000+ employees strong, we're building a company designed to shape the future of private markets for decades to come.

Our culture is built for people who want to do ambitious, meaningful work alongside exceptionally talented teammates. We think like owners, move with urgency, and take pride in solving hard problems that truly matter to our customers and the future of private markets. We believe the best ideas come from open debate, deep collaboration, and diverse perspectives, which is why we believe transparency is the default and feedback makes us stronger. If you're energized by high standards, rapid growth, and the opportunity to help define a category at a pivotal moment, come join us!

Juniper Square offers employees a variety of ways to work, ranging from a fully remote experience to working full-time in one of our physical offices. We invest heavily in digital-first operations, allowing our teams to collaborate effectively across 27 U.S. states, 2 Canadian Provinces, India, Luxembourg, and England. We also have physical offices in San Francisco, New York City, Mumbai and Bangalore for employees who prefer to work in an office some or all of the time.

About your role

We are looking for a Lead Product Manager to own the agent runtime and evaluation layer of Foundry, the platform where Juniper Square builds, deploys, and operates AI agents, apps, and code. It is built for non-engineers: a fund administration specialist or business SME should be able to describe what they need and end up with a governed, deployable automation that has its own identity, access grants, and audit trail. You will report to the Group Product Manager for Foundry and work closely with a dedicated engineering team on the details that make that promise real.

About 75% of your time goes to the agent runtime itself: versions and deployments, evaluations and testing, and what happens when a new model ships and teams want it for some automations but not others. The other 25% goes to partnering with the workflow orchestration PM on the contracts that let hosted agents plug into governed workflows and get real work done. Think of orchestration as the shape of the container, and this role as how that container gets filled with automations in a deterministic, predictable way.

The platform is early and the problems are hard. We are rebuilding how fund administration gets done to be agentic first, with humans and agents building, configuring, and deploying automations together. You will help shape the foundations, and engineering will look to you to make the call on ambiguous platform tradeoffs.

AI fluency is a baseline expectation. PMs at this level should be comfortable using the full range of modern tools (e.g. LLMs, agent orchestration frameworks, Claude) to move faster and make sharper decisions. That might mean building and deploying your own agents on the platform you own, prototyping before writing requirements, or using coding agents to submit pull requests for engineering. The expectation is sound judgment about what will get to the best outcome for the customer and the team.

This role would benefit from in-person collaboration in San Francisco, where many of our product management, design, and engineering personnel are based, although we also welcome exceptional candidates from elsewhere in the United States and Canada.

What you'll do

• Build the agent runtime for non-engineers: Own the vision, strategy, and roadmap for a platform where business users take an agent, app, or script from idea to governed production without an engineering cycle. Define the full lifecycle: draft, test, publish, version, deploy, run, observe, improve, and retire.
• Own evaluations and model governance: Define how every automation is tested before it ships. Work with engineering and our shared eval platform team on golden datasets, data labeling, and regression gates that block promotion. Decide when an automated judge is enough and when a human has to stay in the loop. Shape how evals decide which models are eligible for which agents, so a new model release becomes a product decision rather than a scramble.
• Solve the \"who evaluates the evaluator\" problem: Agents that grade other agents need their own checks. Partner with engineering on patterns that ke