At ATOMBUILD, I designed a clickstream pipeline for Sony’s OTT data platform that processes 25+ GB every 30 minutes across 600+ event types. It transforms and enriches data for 36+ downstream tables, including with WURFL device and geolocation information.
I also developed observability capabilities for the platform, including schema-drift detection, data-quality checks, and data-volume analysis surfaced through Apache Superset. I optimized Spark jobs, reducing one pipeline’s runtime from about five minutes to three.
For ATOMBUILD’s internal AtomHub platform, I designed reusable, configuration-driven ingestion for databases, Snowflake, Elasticsearch, and cloud services. I implemented near-real-time CDC with Debezium and Apache Kafka, synchronizing PostgreSQL changes with Apache Iceberg.
I built schedulable Iceberg maintenance jobs that improved read performance by approximately 30–40% across client workloads. Earlier, I built PySpark ETL proof-of-concepts at DATAMARCOS and developed SQL queries, stored procedures, and validation logic at TUDIP TECHNOLOGIES.

