At Antenna, I helped modernize 30 ETL pipelines processing 1 TB of data per day across 15 sources, moving them into Python microservices on Google Cloud. The work reduced end-to-end processing time from 14 hours to 2.
I designed architecture across Cloud Storage, BigQuery, Cloud Run, and GKE, and built a Pub/Sub and Dataflow streaming path with Apache Beam for partner event feeds. I also productionized Campaign Lift with data science, replacing notebooks with Python services and dbt models to reduce analysis turnaround from 2–3 days to 1–2 hours.
At Teradata, I built Python and FastAPI recommendation services handling 50,000 requests a day for three internal applications. Precomputed features, PostgreSQL query tuning, and caching reduced p95 response latency from 900 ms to 250 ms.
At Verimatrix, I built Python and PySpark pipelines for security and usage logs, and implemented Kafka ingestion with Spark Structured Streaming. This work reduced median security-event detection latency from 90 to 10 minutes.
