At Turquoise, I rebuilt payer rate ingestion in Go and Python, cutting the nightly backlog from 7 hours to 3 during monthly refreshes. I also streamlined hospital file parsing, reducing peak memory per worker from 11 GB to 4 GB.
I improved Turquoise's pricing analytics by tuning PostgreSQL query plans and partition pruning, bringing p95 requests down from 2.1 seconds to 640 ms on commonly used filters. I also instrumented distributed processors to uncover network retries delaying analytics.
At Urbint, I delivered utility risk feeds using Go gRPC services and Kafka consumers, cutting peak dispatch alert lag from 14 to 6 minutes in pilots. I refactored Python risk processing into parallel partitions, reducing daily scoring from 4 hours to 2 during high-volume periods.
Earlier, I developed telehealth intake services at Ro and transaction processing services at Capital One. Across these roles, I worked on reliable queues, database performance, and production systems in healthcare and financial services.
