Product Engineer

Cognition Cognition · Coding AI · San Francisco, CA · Research & Development

Product Engineer at an applied AI lab building end-to-end software agents, specifically focusing on the Devin AI software engineer and Windsurf AI-native IDE. The role involves owning features end-to-end, building performant and beautiful experiences across IDE, web, and CLI surfaces, setting the standard for agent UX, and working directly on the core agent infrastructure including the agent loop and tool use.

What you'd actually do

  1. Own features end to end, from the first line of code to the conversation with the developer who uses it. User feedback directly shapes what you build next.
  2. Craft interfaces across our IDE, web, and CLI surfaces that feel fast, focused, and effortless. Quality is not a final pass; it's built into how you work.
  3. Define interaction patterns for how developers work alongside AI agents. This is new territory with no established playbook. You'll create the conventions that others will follow.
  4. Work directly on Devin and the core agent infrastructure that powers it. This means building the agent loop, tool use, and the harness that lets Devin operate reliably across real-world tasks.
  5. Ship across our IDE, web, and CLI products. Context-switch quickly and maintain a high quality bar across all of them simultaneously.

Skills

Required

  • Python
  • TypeScript
  • React
  • Full-stack engineering depth
  • Independent product judgment
  • UX and Design Obsession

Nice to have

  • Developer empathy
  • Developer tools experience
  • Relevant industry experience

What the JD emphasized

  • Own features end to end
  • Build the Agent and Agent Harness
  • Move Fast Across Surfaces
  • Strong engineering skills with a natural instinct for product quality
  • Genuine understanding of developer workflows, frustrations, and what delightful tooling actually feels like.
  • Experience with all three in complex, production-grade products.
  • The ability to make sound product decisions without needing extensive process or sign-off.
  • You sweat the details others overlook: micro-interactions, transitions, response times, empty states, because you know that magical product experiences are built from hundreds of small decisions done right.
  • Experience building developer tools, IDEs, or technical products where the user is an engineer is a strong plus.

Other signals

  • building end-to-end software agents
  • AI-native IDE
  • AI that doesn't just assist engineers, but works alongside them as a genuine teammate
  • building the agent loop, tool use, and the harness that lets Devin operate reliably across real-world tasks