At Sei AI, I lead the design and deployment of production LLM and Voice AI systems for enterprise customers across banking, lending, insurance, collections, and fintech. I helped scale the voice-agent platform from 25,000 to 1 million dials per month while improving reliability, latency, and operational tooling.
I built real-time voice agents and multi-agent orchestration with tool calling, memory, transfer logic, and multilingual flows. I also built a pre-release evaluation pipeline that tests prompt, tool, and model changes against simulated customer conversations before they reach live calls.
I redesigned Sei AI’s core document-intelligence algorithm, reducing processing cost per document from $60 to $15 and turning a previously loss-making product into a profitable offering. For mortgage workflows, I worked on extraction, classification, RAG, agentic reasoning, and deterministic policy execution.
Before Sei AI, I led engineering for PollX / MoonSwipeAI from discovery and prototyping through deployment; its visual-AI ad optimization product reduced CPC by 35%. At Walmart Global Tech, I built backend and API features for internal product workflows and contributed to the architecture of a data collection and aggregation application.
