At Umbrella, I built the LLM capability layer for an AI-driven marketing analytics platform, with provider fallback, cost and latency controls, guardrails, and observability. I also implemented multi-agent workflows using live campaign, lead, and budget data, with human approval for AI-initiated writes.
At N.O.A.H., I built crop-yield prediction from satellite imagery and shipped a document-processing pipeline for scanned insurance forms. I also designed RAG and context management for assistants and delivered ML infrastructure and the full stack from prototype to pilot-ready product.
At GdeSlon / KIT Global, I scaled an e-commerce affiliate platform from zero to billions of click-tracking events per day and led its migration from a Rails monolith to services. I grew the team while continuing to write core code, tune performance, and debug production incidents.

