I've led production AI systems at Cognitiv, Brightloom, and Zulily, spanning real-time advertising, restaurant analytics, search, personalization, and recommendations. At Cognitiv, I deployed an AI-powered DSP processing 12B+ daily bid requests with 99.99% availability and a 20 ms online decision-latency target.
I built Cognitiv's agentic orchestration layer with LangGraph and MCP, along with ContextGPT multimodal pipelines and AudienceGPT embeddings that improved purchase-intent prediction accuracy by 22%. My CTR and conversion models improved click prediction by 19% and conversion prediction by 24%, while bid optimization increased return on ad spend by 31%.
At Brightloom, I led an AI/data team building an LLM-powered restaurant analytics assistant that lets users query transactions, diner behavior, campaign performance, and segmentation data in natural language. I also built multimodal semantic-search pipelines across more than 4 million menu images and promotional creatives, and reduced cloud infrastructure spend by 30%+.
Previously at Zulily, I delivered recommendation and ranking services for 15M+ daily users and 10M+ daily search requests. I hold a Ph.D. in Electrical Engineering and Computer Science from the University of Washington and bring deep experience in LLMs, agentic AI, computer vision, distributed machine learning, and production MLOps.
