Kratik Mehta

Meet Kratik

0 following

I build Databricks and Snowflake data pipelines that don't break silently. If your ETL jobs fail without anyone noticing, your warehouse refresh runs all night, or nobody trusts the numbers in your dashboards, that's what I fix.

I'm a Senior Data Engineer with 3+ years building production ETL/ELT pipelines, Lakehouse architectures, and data-quality frameworks for pharmaceutical and clinical-trial data, where accuracy and compliance are non-negotiable.

Recent results:
• Cut a Snowflake data mart refresh from 4 hours to 85 minutes (virtual warehouses + clustering keys)
• Reduced ingestion from 2 hours to 15 minutes with PySpark incremental loading on Databricks and Snowpipe Streaming
• Migrated a 2.97M rows/day production ETL pipeline to an orchestrated platform with quality gates and zero-downtime deployments
• Designed a data quality framework with 114+ validation rules, keeping false-positive alerts below 5%
• Built a scraping pipeline for 900+ California school districts (250,000+ records into PostgreSQL), cutting turnaround from 90 days to 10

What I build:
• Databricks Lakehouse: Medallion architecture, Delta Lake, Delta Live Tables, Auto Loader, Unity Catalog
• Snowflake warehousing: star-schema data marts, Snowpipe, Streams & Tasks, cost and query tuning
• ETL/ELT pipelines in Python, SQL, and dbt from APIs, files, databases, S3, and ADLS
• Data quality and monitoring: validation before and after load, alerting that doesn't cry wolf
• Pipeline rescue: slow Spark jobs, long refreshes, flaky schedules, undocumented code

Stack: Databricks · Snowflake · PySpark · Spark SQL · dbt · Python · SQL · Delta Lake · PostgreSQL · AWS · Azure · Git

Certified in Databricks and Snowflake.

Send me your data sources and what's not working, and I'll reply with how I'd approach it.