Data Engineer LTM (Formerly LTIMindtree)
Designed and automated daily production ETL pipelines for the RDNA 2.0 P&G AMA Multi-Publisher Platform using Python, Apache Airflow, and GCP Cloud Composer. Built publisher-specific parallel TaskGroups to ingest more than 10 TB of campaign data from Meta, TikTok, Snapchat, DV360, and Google Ads across 17+ markets, reducing DAG runtime to approximately 40 minutes.
Engineered a two-layer BigQuery lakehouse with Silver materialized views and Gold star-schema tables. Applied partitioning and clustering across time, market, and brand dimensions, reducing query latency by 40% and lowering scan costs.
Established additive and non-additive metric taxonomies to prevent cross-publisher double counting, including unique-reach aggregation, frequency derivation, spend normalization, and daily FX conversions into USD. Enabled Databricks Unity Catalog JDBC federation for zero-copy access to BigQuery data.
Owned Looker Studio reporting for three regions and 26 countries, improving dashboard render speed by 30%. Led production issue remediation through SQL window-function deduplication, NULL-safe joins, and custom rollup logic for complex BI aggregation requirements.