At Sydon AI, I built an external-facing MCP Operations Plane with 17 tools and 11 read adapters connecting AI agents to enterprise data and APIs. It uses tenant- and account-scoped access, fail-closed policy filtering, data minimization, and human-approved write proposals.
I also developed an AI Listing Builder agent with scheduled reviews, keyword analysis, and approval workflows. I optimized processing from 400 ms to 5.5 ms and processed 9,231 listings in 7.4 minutes.
At EQ App, I built multi-agent workflows with LangGraph and CrewAI for resume parsing and initial qualification, cutting manual screening time by 60%. I integrated LinkedIn, Zoom, and an ATS through REST APIs for sourcing and scheduling.
At CodeSizzler, I delivered enterprise AI apps and built custom RAG pipelines that improved retrieval relevance by 40%. In my voice-agent project, I built an LLM agent for inbound calls with RAG, call routing, and live-agent handoff fallback.

