Description
We're building a production agentic AI system for an enterprise client in financial services. This is real production work, not prototyping — agents that support business users by reasoning over governed enterprise data and taking actions across internal tools. You'll be building role-specific agents and the orchestration that connects them, working from an architecture and design that's already defined. We need someone who can execute against a clear spec with high quality, not a strategist. What you'll do: Build multi-agent workflows using LangGraph (or equivalent orchestration) Implement retrieval over enterprise data (RAG) and tool/function-calling Integrate agents with data through a governed access layer — agents never touch raw database credentials directly Work with Snowflake Cortex (Analyst / Search) and the Model Context Protocol (MCP) Build evaluation and quality checks into every agent you ship Required: 3+ years building production LLM applications (portfolio or GitHub required — not just demos) Strong Python Hands-on LangGraph, LlamaIndex, or comparable agent framework RAG, tool-calling, and prompt engineering in production Experience with governed / API-mediated data access, not credential-dumping You build evals as a matter of habit Also important: Snowflake Cortex, AWS Bedrock, or Azure AI Foundry MCP / agent-to-agent (A2A) patterns Financial services or other regulated-industry experience Knowledge graphs / semantic layers Please include: a short note on a p