CFA Institute, Remote, US
RAG pipelines – Delivered retrieval-augmented generation workflows on top of enterprise content, including chunking, vector indexing (Azure Blob), and golden-set evaluation harnesses to measure accuracy, latency, and cost per query.
MCP server – Built a Model Context Protocol server over existing OpenAPI 3.0 endpoints, exposing safe, authenticated, rate-limited tools to GitHub Copilot and other MCP-aware clients.
Custom VS Code integration – Developed a Local Copilot Protocol (LCP) extension providing a domain-specific entry point in the IDE, reducing context switching and accelerating prototyping for internal developers.
OpenWebUI deployment & extensions – Deployed and customized OpenWebUI with observability hooks, deterministic fallbacks, and prompt versioning to support controlled experimentation with LLM-driven features.
Governance & DX focus – Ensured all AI-enabled features were observable, cost-controlled, and compliant, while packaging SDKs and dev tools that cut time-to-prototype by 40%+.