At SynthralOS, I architected and shipped an AI-native workflow automation platform from zero to production. I developed a natural-language-to-workflow compiler that turns prompts into runnable node graphs.
I unified 11 LLM providers and 12+ agent frameworks behind a vendor-agnostic router, with per-run cost tracking and observability. I also developed persistent RAG memory using pgvector, Qdrant, tiered memory, and a knowledge graph.
At Paramet, I build AI assistants that expose application capabilities as callable tools, and developed an MCP server backed by a relational database. I also automated CAD workflows through the Autodesk Inventor API, reducing design cycles from multiple days to under 30 minutes.
My projects include a real-time voice agent connected to CRM, calendar, and email services, plus AWS-based legal document and company knowledge assistants. I work across Python, LLM orchestration, RAG, and cloud AI services.
