At Event Manager, I designed an LLM matching engine that turns event briefs into venue proposals using SQL filters, pgvector semantic ranking, and validated outputs. I also built a web-extraction pipeline that reduced page content by over 99% and verified source quotes in code.
At CME Group, I led a six-person team and owned pipeline delivery. I cut a daily identity-matching pipeline across Salesforce and trading data from 5–6 hours to 10–15 minutes by moving fuzzy matching into BigQuery SQL and reducing candidate pairs with blocking filters.
At BOMBORA, I built ad-metrics pipelines into BigQuery and client platforms, and designed a PostgreSQL job-state store that stopped duplicate Dataflow runs. I also built containerized Java REST services on AWS and added Datadog monitoring.
Earlier, at AdRoll, I implemented MapReduce audience and identity-graph jobs and worked on real-time Storm pipelines. My work spans data engineering for ad-tech and financial platforms, alongside recent LLM applications and research projects in cryptography and zero-knowledge proofs.
