I've built production AI systems at Sekovia.app, including a voice AI agent combining Claude reasoning, real-time STT/TTS, and Twilio telephony at 1.5s p50 latency and $0.13 per call.
At iTradeNetwork, I developed an LLM-powered normalization pipeline that processed 16M+ transactions in under four minutes and saved 1,788 analyst hours annually. I also cleared a 50K-record mapping backlog through LoRA fine-tuning and built agentic workflows for supply-chain data normalization.
I've also delivered RAG and agentic AI products at LexLead.ai and the University of Chicago Data Science Institute, and used Python, SQL, Tableau, and Azure Databricks to improve Sephora member engagement at Digitas.
