At Stitch Fix, I own the shared ML platform behind 40+ production models across recommendations, forecasting, merchandising, and generative AI. I built personalized retrieval and ranking for stylist recommendations and production LLM services for stylist assistance and product-content generation.
I design Kubernetes-based model-serving systems on AWS, using Terraform-managed infrastructure, containers, batching, caching, autoscaling, and workload-aware routing. This reduced p95 inference latency by 43% and serving cost by 36%, while monitoring, evaluation, canary releases, and automated rollback keep production AI reliable.
Previously at Thumbtack, Wayfair, and Zillow, I built matching, ranking, search, recommendation, NLP, and feature-pipeline systems operating at marketplace and catalog scale. I enjoy taking models from raw events through curated features and deployment to measurable customer outcomes.
