At Goava, I own the core data platform behind a B2B sales intelligence product, which processes 4M company and contact records. I built LLM-based extraction pipelines that improved accuracy from 75% to 95% over the previous rule-based approach.
I also automated open-source LLM inference on EC2 Spot instances, cutting inference cost by 66%. For event-driven processing, I architected ECS and SQS workloads handling 1M messages per day.
Earlier at Goava, I rearchitected the company data ETL pipeline, reducing extraction and processing time by 85% across 4M company documents. I designed Athena and Iceberg partitioning and bucketing strategies to cut analytical query cost and runtime.
I progressed from Junior Software Engineer to Senior Software Engineer on the same platform. My work has included distributed crawling with Apache Nutch, a NER pipeline processing 300k news articles per day, and data ingestion from external sources into an S3 data lake.

