ML & Agentic Systems Engineer [IC4] - Sourcegraph

Who we are

Sourcegraph is on a mission to bring clarity and control to the world's most complex codebases. As agentic development becomes the dominant engineering paradigm, we provide the context layer teams need to take control of their codebase. We are a globally distributed team that values high agency, direct communication, and customer love.

Why this job is exciting

The Code Understanding team owns the surfaces where intelligence meets the developer: Deep Search, Query Assist, Smart Hovers, and the APIs that both humans and AI agents rely on. As a Staff ML and Agent Engineer, you will be the technical owner of our model strategy, evaluations, and agentic systems.

What you'll do

• Agentic systems: Design and harden multi-step, tool-using agent loops for enterprise-scale products.
• Evaluations: Build representative datasets, baselines, and error taxonomies to measure model performance and agent behavior.
• Model lifecycle: Own model selection, upgrades, and fine-tuning to optimize for quality, cost, and latency.
• Context engineering: Push the boundaries of how we ground models in customer code via retrieval, ranking, and context window management.
• Technical leadership: Set standards for the team, mentor engineers, and influence technical direction beyond your immediate scope.

About you

• Production ML experience: You have personally owned a production model lifecycle from dataset construction to monitoring.
• Agentic expertise: You have built reliable, observable, and cost-bounded multi-step agentic systems.
• Evaluation judgment: You understand how to connect offline metrics to production behavior.
• Pragmatic engineering: You prioritize simplicity and robustness, treating cost and latency as core product constraints.
• Technical stack: You are a strong software engineer comfortable with Go, TypeScript, GraphQL, and Docker.
• Autonomy: You thrive in ambiguous environments, turning rough product ideas into concrete milestones.

Hours & location

While we hire almost anywhere, we have a preference for candidates in North America or Europe. Working hours must overlap with EST for at least 20 hours/week.