VIJAY SAGI

About

Data/ML platform engineer with ~6 years in financial services and enterprise data platforms, building regulated fraud/credit-risk pipelines on AWS (Airflow, EMR/Spark, MLflow) and an Azure lakehouse ingesting 50+ sources and serving ~100M records/day. Also led an early-stage deployment of a RAG-based internal AI search tool using AWS Bedrock and LangChain with automated evaluation to validate LLM accuracy.

Experience

Education

Key Strengths

• Owned end-to-end regulated financial data pipelines (ingestion → validation → transformation → serving) for fraud detection and credit risk modeling
• Built automated workflows with Airflow and Python; implemented early data-quality KPIs to catch schema drift/missing values
• Scaled Spark/EMR ETL processing ~2M–5M records with SQL optimization and Spark tuning
• Refactored legacy batch jobs into modular, observable pipelines improving reliability and maintainability
• Integrated MLflow for experiment tracking/model versioning to keep training/serving feature consistency
• Reduced manual review by 30% via automated feature validation and ML tracking
• Architected Azure lakehouse ingesting 50+ enterprise sources and serving ~100M records/day to internal teams
• Improved batch processing performance by ~70% using partition optimization, incremental loading, and standardized IaC/CI-CD (Terraform)
• Handled schema evolution and fault tolerance using Delta Lake features, structured streaming, and checkpointing
• Led end-to-end deployment of a RAG-based AI search tool using AWS Bedrock + LangChain with automated LLM evaluation