Srishti Shetty

Engineered an AWS Glue + Amazon Redshift governed data integration architecture consolidating 12 healthcare data sources into a warehouse used daily by 300 stakeholders. Authored BRDs/FRDs in Confluence/JIRA with acceptance criteria and use cases that reduced UAT cycles from 21 to 5 days. Architected an enterprise RAG pipeline on Amazon Bedrock and LangChain, using LLM fine-tuning via Amazon SageMaker to automate clinical data extraction and eliminate 240 review hours per quarter. Built Power BI executive KPI scorecards using advanced DAX, compressing reporting cycles from 3 weeks to faster cadence. Deployed Python predictive modeling integrated with Snowflake to improve supply chain inventory forecasting across 1,400 distribution centers and prevent $300K+ annual losses. Optimized complex SQL workloads on Databricks Delta Lake, reducing pipeline runtime from 4 hours to 18 minutes for a 50M-record healthcare dataset.

Processed, validated, and reconciled 40K+ FOIL records using SQL and Excel, enforcing 12 data integrity checks for consistent, audit-ready reporting across six housing compliance workflows. Designed and deployed 5 interactive Power BI dashboards to surface multi-year trends and reduce reporting turnaround from 5 days to 4 days. Automated document parsing and data preparation workflows with Python and VBA, eliminating 15+ hours of manual work per week across five teams. Collaborated with program and compliance stakeholders to define data quality standards, translating regulatory requirements into eight structured data pipelines and 20+ reusable SQL queries. Consolidated inspection and permit data from three source systems into a unified SQL reporting layer, enabling discrepancy resolution in under 2 hours instead of 2 business days.

Designed a real-time anomaly detection and statistical analysis model in Python and Apache Kafka to flag $100K+ fraudulent payment patterns across 2.4M daily transactions within 30 days of launch. Built Tableau fraud analytics dashboards integrated with PostgreSQL, enabling the Risk team to investigate 150 cases daily and cut mean resolution time by 40 minutes. Analyzed A/B testing outcomes and cohort behavior across 500K monthly active users in Python and SQL, identifying three checkout funnel drop-off patterns and reducing weekly abandonment from 18,000 to 11,000. Synthesized transaction data from six sources into bi-weekly executive KPI reports using Tableau and SQL to track $12M daily transaction flows across eight payment corridors.