Jathin Darsi

Data Scientist @Blue Cross Blue Shield of Michigan

TX, US

Cut compliance documentation cycle time by 30% and improved transcription accuracy by reducing Word Error Rate (WER) from 12% → 8%(33% improvement) through deployment of a HIPAA-compliant Whisper + GPT-4 + LangChain pipeline on Azure Functions.• Reduced claim adjudication latency by 25% by architecting multi-agent decision engines (LangGraph + CrewAI) on Azure with fallback logic and memory routing, while increasing retrieval Precision@5 by 18% for policy lookups.• Enhanced fraud detection pipelines in Azure Machine Learning, boosting F1-score from 0.72 → 0.85 and raising anomaly detection precision on high-risk claims by 20%.• Aligned AI solutions with HL7/FHIR standards and ensured audit-ready HIPAA/GDPR compliance, reducing compliance audit failures to <2%.• Trained and educated 20+ compliance and IT staff on responsible AI practices, fostering adoption while reducing model hallucination rate by 15%(LangSmith evaluation + Azure Monitor dashboards).• Optimized ML pipelines in Azure Databricks and deployed monitoring with Azure Monitor + Jaeger, cutting latency by 15% and improving issue resolution speed by 20%.