Who we are
Everything is changing in how software gets built, and Sourcegraph is at the center of that transformation. With Code Search, Deep Search, and MCP, Sourcegraph is the world’s most powerful code intelligence platform that developers and agents rely on to navigate, understand, and operate on massive, complex codebases with speed and confidence.
Teams at companies like Stripe, Uber, and Dropbox rely on Sourcegraph to ship faster and with higher quality. We’re backed by a16z, Sequoia, and Redpoint, and proud to operate as a globally distributed team that values high agency, direct communication, and a deep love for developers and their craft.
Why this job is exciting
The Code Understanding team owns the surfaces where intelligence meets the developer, including Deep Search, Query Assist, and Smart Hovers. As a senior agent engineer, you will be the technical owner of our agentic systems, setting the direction for the team and making our products faster, cheaper, and more accurate.
Concretely, you'll own work like:
- Agentic systems: Design and harden multi-step, tool-using agent loops at enterprise scale.
- Pragmatic use of evaluations: Develop judgment on where evaluations earn their keep and how to use metrics to avoid noise.
- Models: Manage model selection, upgrades, and fine-tuning.
- Retrieval and context engineering: Push the boundaries of grounding models in customer code through improved retrieval and ranking.
- Cost and latency: Treat efficiency as a product feature by profiling and optimizing model performance.
About you
You are a senior engineer and technical leader with hard-won skills in agent engineering, blending software engineering, machine learning, and statistics. You operate at staff scope, owning ambiguous problems and setting standards for the team.
- Agent engineering depth: Experience building, training, and operating models in production.
- Agentic fluency: Ability to design multi-step agentic systems that are reliable and observable.
- Pragmatic mindset: Focus on real-world feedback and lightweight testing over unnecessary perfectionism.
- Autonomous operation: Ability to translate ambiguous requirements into prototypes and actionable milestones.
- Technical breadth: Proficiency in our stack (Go, TypeScript, GraphQL, Postgres, Docker) or the ability to master it quickly.