Ayush Kumar
Sahu

Data Engineer | ByteIQ Analytics

Jan 2026 – Present | Remote

• Designed and built the ingestion layer for a global US relocation services company's SQL Server to AWS warehouse migration (operational and reporting sources), helping shape the target architecture and proposing Airflow for orchestration.
• Engineered a CDC pipeline on SQL Server with AWS DMS, running full load plus ongoing replication into an S3 bronze layer as partitioned Parquet, delivered with a least-privilege reader, runbook and security report.
• Developed a Python CDC extractor with pyodbc and pyarrow that reads SQL Server change tables by LSN range via fn_cdc_get_all_changes, writing Parquet in the AWS DMS S3 folder layout.
• Created a config-driven extraction framework where a YAML manifest maps each table to one of four strategies (rowversion, datetime, identity and snapshot hash), with per-table control state for resumable loads.
• Automated loading with Apache Airflow across two DAGs and seven tasks, using an idempotent file-level control table and validation that fails the run on any row-count mismatch between staging and the control log.
• Merged two disjoint source systems (53,860 rows across 24 tables) into a unified SQL Server warehouse with 18 dbt models, recommending hash-based surrogate keys after analysis showed no shared keys.
• Reverse-engineered the client's legacy warehouse, mapping its hub-centred star schema (30 dimensions, 27 fact tables, 95 stored procedures), and authored a 20 section platform reference plus onboarding guide.