Title: Specialist Data Engineering
Location: Pune
Exp: 4-7 Years

Key Responsibilities
Architect and implement the three-tier data storage model: Azure SQL (live transactional SNPD DB) ADLS Gen2 Delta Lake (historical analytics / ML features, nightly ADF ETL) Azure AI Search (vector / RAG index).
Build and maintain Azure Data Factory (ADF) pipelines: nightly CDC-based ETL from the SNPD SQL database, SAP (cost/PO/BOM/vendor masked at API layer), and Teamcenter PLM (BOM snapshots, part lifecycle, ECN).
Design and implement the unified SNPD data model: enforce project_id + part_number as universal pivot keys across all tables; ensure referential integrity across SNPD core domain tables and migrated portal tables (NVPC, RFQ, PPAP, CDMM, etc.).
Build the AI feature store on Delta Lake: dl_gate_cycle_times, dl_supplier_risk, dl_nvpc_benchmarks, dl_cost_variance, dl_deliverable_actuals with incremental refresh, partitioning, and Z-ordering for query performance.
Conduct data audits on SAP S/4 HANA, Teamcenter PLM, and all 8 legacy homegrown portal databases; assess data quality,



identify gaps, and remediate for ML readiness.
Implement SAP cost data masking at the API / pipeline layer sensitive pricing data must be obfuscated before reaching any MCP server or AI agent.
Set up Azure AI Search vector index: embedding ingestion pipeline from the document store (SharePoint / Azure Blob), chunking strategy, metadata schema, and incremental re-indexing on document updates.
Establish data lineage, quality checks, and observability: row counts, null rates, schema drift alerts, and SLA monitoring for all ETL pipelines.
Support historical data migration: 5 7 years of legacy SNPD and portal data into the unified SNPD database; validate referential integrity and completeness post-migration.
Collaborate with the ML Engineer to serve training datasets from Delta Lake; optimize feature computation using Synapse Serverless or Databricks as compute.
Implement RBAC and data acce

📌 Specialist Data Engineering (Pune)
🏢 NR Consulting
📍 Pune