SHAIVY MITTAL

Project IRIS  AI-Powered Analytics & Executive Summaries (Top Impact)

• Designed and launched an AI summarization pipeline for 200+ WBR/DBR reports consumed by L8+ leaders across ~15 AWS services.
• Built a Python Lambda summarizer invoking AWS Bedrock LLMs and integrated it into existing report-generation flows.
• Created a prompt library, tuned model context windows, and executed end-to-end accuracy testing to finalize production prompts.
• Impact: Reduced PM analysis + summary creation effort by ~90%, saving 5-6 hours per PM per week across services.

SageMaker Unified Studio - Cross-Service Data Platform

• Onboarded data for a 2024-launched service; ingested operational, revenue, and clickstream signals into a multi-layer S3 data lake.
• Built PySpark Glue ETL pipelines (CDK-deployed) supporting ingestion, schema evolution, validation, partitioning, and $200k influenced revenue attribution across 8 AWS services.
• Developed QuickSight dashboards, JSON reports, and AI-assisted analysis scripts, reducing PM analysis workload by 4-5 hours weekly.
• Leveraged Cedric and Quicksuite to generate automated customer funnel, usage, and revenue stories cutting PM effort by ~50%.

• Refactored telemetry pipelines; redesigned 60+ datasets, 8 DBRs, and 4 WBRs using S3 compaction, partition pruning, and Parquet optimization.
• Achieved 83% query runtime reduction through CDK-driven infra improvements, Redshift materialized views, and SQL rewrites.
• Built cross-service revenue attribution ETLs using PySpark Glue, reducing analyst manual work by 4060%.
• Created a reusable Glue + Athena ETL framework for standardized ingestion and metric computation, cutting investigation time by 50%.
• Designed EMR usage + revenue ETLs and a QuickSight dashboard enabling leadership insights and reducing PM analysis by 4-5 hours weekly.

• Built ETL pipelines using Informatica, Oracle SQL, and Python; automated data validation frameworks.
• Migrated OBIEE reporting to SAP Concur and improved ETL scheduling performance.