Design and build agentic systems — multi-step agents that plan, call tools, retrieve context, and take action with appropriate human-in-the-loop checkpoints
Build MCP servers and clients to securely expose client data, internal tools, and APIs to LLMs in a standardized, auditable way
Ship LLM-powered applications: copilots, document intelligence, search, summarization, and workflow automation
Design and maintain RAG pipelines — chunking, embeddings, vector stores, retrieval, reranking, and grounding
Integrate model APIs (OpenAI, Anthropic, Bedrock, Azure OpenAI, open-weight models) and pick the right model for the job based on quality, latency, and cost
Develop evals and observability for agents and AI features so we know what's working in production and what's regressing
Apply prompt engineering, structured outputs, function/tool calling, and guardrails to make agent behavior predictable
Write production Python backends and APIs that expose AI capabilities to web and mobile clients
Collaborate with engineers, designers, and product folks to scope what AI should (and shouldn't) do in a given product
Help shape responsible AI practices for federal use — privacy, security, auditability, and human oversight
Requirements:
5+ years of professional software engineering experience, with at least 1 year shipping LLM-based or AI-powered features to production
Hands-on experience designing or building agentic systems — tool calling, multi-step reasoning, planning loops, or agent orchestration (LangGraph, CrewAI, OpenAI Agents SDK, Claude tool use, or equivalent)
Working knowledge of the Model Context Protocol (MCP) — or demonstrated ability to pick it up quickly, plus familiarity with the broader landscape of agent/tool standards
Strong Python and experience building and deploying backend services and APIs (FastAPI, Flask, or similar)
Hands-on experience with at least one major LLM provider (OpenAI, Anthropic, Bedrock, Azure OpenAI, Vertex, or open-weight models via vLLM/Ollama)
Working knowledge of RAG: embeddings, vector databases (pgvector, Pinecone, Weaviate, Qdrant, or similar), and retrieval evaluation
Comfort with prompt engineering, structured outputs (JSON mode, schemas), and tool/function calling
Experience writing evals — even lightweight ones — for non-deterministic systems
Solid SQL and experience with relational and unstructured data
Familiarity with at least one cloud platform (AWS, Azure, or GCP)
Git, code review, and modern collaborative workflows
Strong written and verbal communication — you can explain AI tradeoffs to non-technical stakeholders.
Benefits:
Competitive pay
Contribution toward health benefits
Work from anywhere in the US
High-visibility federal projects with real impact
Small team where your ideas actually ship
Generous exposure to the latest AI tooling and models