About the role
Manage the design and delivery of multi-agent systems. You'll architect agent orchestration, author MCP servers, design RAG systems, and set the engineering standards the team follows.
Responsibilities:
- Design and implement multi-agent orchestration using LangGraph (state, checkpointing, human-in-the-loop).
- Author MCP servers - tool schemas, auth, error contracts, versioning.
- Architect RAG systems - retrieval strategy, reranking, hybrid search, evaluation.
- Build and maintain the agent evaluation and observability stack.
- Deploy and operate agents on Azure Foundry Agent Service and Azure Functions.
- Establish agentic patterns and reference implementations for the team.
- Mentor junior engineers and lead technical design reviews.
Required Skills:
- Programming: Python (expert) - async, typing, packaging, performance.
- Framework: LangGraph in production - multi-agent orchestration, state management, checkpointing, human-in-the-loop.
- Framework: LangChain (tools, retrievers, callbacks) and FastAPI (async, auth, streaming).
- Platform: Azure Foundry Agent Service (provisioning, threads, tool wiring) and Azure Functions.
- Skills: AI agentic orchestration and agentic patterns - planner/executor, supervisor, reflection, tool use, multi-agent collaboration.
- Skills: MCP server design and authoring - tool schemas, auth, error contracts, versioning.
- Skills: RAG architecture - retrieval strategy, hybrid search, reranking, evaluation.
- Skills: Prompt engineering with structured eval frameworks (LangSmith, Promptfoo, or equivalent).
- Skills: Agent observability - tracing, App Insights, token/cost telemetry, failure-mode analysis.
- ML / Python libraries: PyTorch, Hugging Face Transformers, sentence-transformers, scikit-learn, NumPy, Pandas.
- Skills: Identity - Okta, OAuth2, JWT (client_id propagation, scopes).
- System design: Multi-tenant SaaS, idempotency, retries, queueing.
Desired Skills:
- Platform: Azure Agent Framework (Python).
- Skills: Fine-tuning (SFT/LoRA), model routing, cost/latency optimization, semantic caching.
- Skills: Pydantic, DSPy, or similar typed/optimized agent frameworks.
- Programming: Angular or React for Agent / Workflow Studio surfaces.
- Platform: Azure DevOps or GitHub Actions, Bicep / Terraform IaC.
- Data: PostgreSQL, SQL Server, semantic layer / dataset definition design.
Experience:
- 5+ years building software in Python.
- 2+ years building production LLM or agentic systems.
- Track record of leading technical design and mentoring engineers.