Software Engineer (Agent Platform) - Intern - 2026-2027
On-siteSan Francisco, California, United States
Job Summary
Design and ship the orchestration layer that routes, sequences, and supervises multi-agent workflows across customer environments in real time. Build tooling, sandboxes, and eval loops enabling agents to test and validate outputs before production deployment. Contribute to the Agent Manager system where agents author, evaluate, and optimize other agents autonomously using guardrails and human-in-the-loop checkpoints. Design, code, and ship full-stack features from data models to APIs and front-end tooling while co-creating with customers to surface trust edge cases. Join a team of builders from Scale, Databricks, and Meta working on frontier AI for the physical economy.
Required Qualifications
- Comfort with Python (asyncio, FastAPI)
- TypeScript (Node, React)
- CS fundamentals in distributed systems
- Coursework, internship, or personal-project experience with LLM APIs, eval frameworks, RAG, or agent frameworks (LangChain, MCP, or similar)
- Interest in how multi-step, multi-agent systems coordinate state, tool calls, and failure recovery
- Track record of shipping full projects (school, internship, or personal) end-to-end
- Comfortable with ambiguity
Desired Qualifications
- Experience with agent frameworks (LangChain, MCP, or similar) — production experience not required, curiosity is
- Experience building multi-agent systems
- Experience with guardrails, rollback paths, and human-in-the-loop checkpoints
- Experience designing and shipping full-stack features for agent platforms
- Experience working with real customer workflows to surface where automated agent-improvement can and can't be trusted
- Experience turning edge cases into platform safeguards
- Experience defining best practices for agents building agents
Hiring someone like this?
Get your role in front of qualified candidates on Sorce.