Senior Software Engineer

About the role and team 

At Uber, AI agents are increasingly writing, reviewing, testing, and debugging code in production. They deliver high-quality results efficiently, without wasting tokens, only when they understand Uber's codebases. The AI Context Engineering team builds the context foundation that every software-development AI agent at Uber relies on:

• AI Context Graph: a unified graph of services, ownership, domain entities, and business flows
• Code Graph: a graph representation of Uber's monorepos built for agents to consume
• AI-Generated Documentation: documentation that stays fresh as code changes
• Agentic Memory: a memory layer that lets agents learn across sessions

As a Senior Software Engineer, you will own major parts of this platform end to end. You'll take them from prototype to production at monorepo scale, and work with partner teams to integrate foundational context products with the agents running in production for workflows such as root-cause analysis, code authoring, test planning, and library upgrades, etc

What will you do

• Context Graph and Ontology: Enrich the graph with new entity types, such as data streams, metrics, and observability signals, and link them to code so agents can reason across systems.
• Code Graph Infrastructure: Build and scale code graphs across multiple monorepos and microrepos, covering symbols, call edges, dependencies, and links between code and tests. Add incremental updates, versioned (historical) graphs, and function-level embeddings, and serve the graphs to agents with low latency through local and remote interfaces.
• AI-Generated Documentation: Extend automated documentation from backend services to mobile (screens, UI components), libraries, and data pipelines. Use refresh strategies that keep context current as code changes.
• Agentic Memory: Design the memory layer that lets agents persist and reuse decisions, past fixes, and team preferences across sessions.
• Standards and Governance: Define and roll out standards for agent instruction files (AGENTS.md), including structure, content, and token budget, and generate them systematically across modules.
• Context Quality Evaluation: Build continuous evals that measure the accuracy and freshness of every context product, and run A/B experiments that prove the impact on agent outcomes.
• Tooling and Integration: Build MCP servers, skills, and tools that deliver context to agents. Partner with AI Platform, Developer Experience, and product teams to integrate them into production workflows.
• Technical Leadership: Drive design reviews, mentor engineers, and turn ambiguous problems into well-scoped, measurable projects

What will you need

• Experience: 8+ years of software engineering experience building production systems, with a degree in Computer Science or a related field, or equivalent practical experience.
• Programming: Strong in Go, Java, or Python, writing clean, well-tested, maintainable code.
• LLMs and Agentic Systems: Hands-on experience building with LLMs and agents, including prompt and context design, tool use, and retrieval, and a good sense of what context makes agents succeed or fail.
• Code Understanding: Experience with static analysis, ASTs, parsers such as tree-sitter, or build and dependency systems in large codebases.
• Data at Scale: Experience building reliable data pipelines and working with graph or other large-scale data stores.
• Emerging Protocols: Familiarity with MCP and agent-to-agent protocols and how to apply them.
• Problem Solving: Strong analytical skills and a habit of measuring outcomes, not just shipping features.

Success Metrics

• More agentic workflows use curated context in production, with a measured improvement in each workflow's business metric, such as accuracy, cost, latency, or CI time.
• Agents provide higher-quality, better-grounded answers while using fewer tokens and tool calls than when they rediscover context from scratch.
• Context coverage grows across languages, repos, and entity types, with measurable accuracy and freshness.

Bonus Points If you have

• Graph Databases: Experience with Neo4j or similar graph databases, including query optimization and schema design.
• Embeddings and RAG: Experience with embedding generation, vector databases, and retrieval pipelines.
• Evaluation: Experience designing LLM evals, LLM-as-judge methods, or A/B experiments for AI systems.
• Developer Platforms: Experience in developer tooling, CI, or monorepo infrastructure.
• Observability: Experience with Prometheus, Grafana, Langfuse, or similar tools for monitoring AI systems.