Software Engineer

Cognition•San Francisco, CA

13d•$260,000 - $300,000

About The Position

We are an applied AI lab building end-to-end software agents. We're the makers of Devin, the first AI software engineer. Our team is extremely talent-dense. Among our founding team, we have world-class competitive programmers, former founders, and leaders from companies at the cutting edge of AI including Scale AI, Palantir, Cursor, Waymo, Tesla, Lunchclub, Modal, Google DeepMind, and Nuro. Building Devin is just the first step—our hardest challenges still lie ahead. If you’re excited to solve some of the world’s biggest problems and build AI that can reason on real-world tasks, apply to join us.

Requirements

Systems engineering depth: Experience building reliable, performant distributed systems; you have strong opinions about correctness, failure modes, and production behavior.
Product instinct: You care about how the software you build feels to use and you have shipped things that real people depend on.
Comfort with ambiguity: You can make progress on hard problems with incomplete specs, learn fast from results, and course-correct without needing a lot of direction.
Velocity without shortcuts: A track record of shipping quickly while maintaining the kind of code quality that a high-density team expects.
Curiosity about agents and AI: You have dug into how LLMs work, how agents fail, and what it takes to make AI-powered systems behave reliably in the real world.
Strong Python proficiency: Python is the primary language across Cognition's codebase; you write clean, well-structured Python and are comfortable owning large Python codebases in production.
Relevant industry experience: Prior experience at a frontier AI lab, applied AI company, or developer tools company; you know what good looks like in this category.
Degree from a top-tier university: BS, MS, or equivalent in Computer Science, Mathematics, Engineering, or a related technical discipline from a highly selective program.

Responsibilities

Build core agent infrastructure: Design and ship the systems that power Devin's long-horizon task execution: tool use, context management, multi-step planning, subagent orchestration, and sandboxed code execution environments.
Improve Windsurf as an AI-native IDE: Contribute to editor intelligence, agent-in-the-loop workflows, real-time code understanding, and the developer experience that makes Windsurf different from every other IDE.
Close the loop between models and products: Work directly with researchers to translate new model capabilities into shipped features; your feedback shapes what gets prioritized in training.
Own reliability and performance at scale: Build systems that handle millions of agentic tasks with low latency, high reliability, and the kind of correctness that developers depend on in production.
Move the category forward: Cognition is defining what AI software engineering looks like. You will have real input into what gets built next and why.