SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Baseten is building the inference platform for AI companies like Cursor, Notion, and Abridge. You'll own the internal AI developer platform end-to-end, creating the infrastructure that makes AI-first software development the default at Baseten rather than an afterthought.
The core challenge: today every engineer assembles their own agent configs, context files, and MCP servers. Good patterns stay trapped in individual setups. You'll build the platform that makes the good path the easy path—agent configurations tuned to the monorepo, context and tooling layers that make agents competent in the codebase, evals that measure what actually works, and rollout mechanics that get new engineers productive with AI in week one.
You'll build the agent substrate (repo-level context infrastructure, internal MCP servers for CI/observability/deployment, shared skills and subagents), the golden path (project templates, self-serve infrastructure, gateway and cost controls), and the feedback loop (eval harnesses, instrumentation of AI tool usage, honest reporting on what works and what doesn't). You'll also integrate agents into the SDLC—PR review triage, test gap-filling, incident context, migrations—with guardrails that make it trustworthy.
This is a platform engineering role, not an AI evangelist position. Success is measured by adoption because your work is better than what teams would build themselves, not by mandate. You'll evaluate and integrate third-party AI coding tools (Claude Code, Cursor, etc.), establish evaluation practices for AI-assisted development, embed with teams to find genuine wins, and own the safety layer (permissions, secrets, audit trails, cost management).
You need 4+ years building AI-native SDLC, strong Python/Go skills, hands-on experience with LLMs and agent frameworks (tool calling, MCP, context management, orchestration), deep personal fluency with AI coding tools, and a platform mindset. Developer tooling, CI/CD, and Kubernetes/Docker fundamentals required. Comfort with ambiguity is essential—this space invalidates its own best practices every few months. Excellent written communication is critical; much of your leverage comes through docs, templates, and examples that scale beyond direct conversations.
