Health Domain

Project 1: Patient Data Lake for Health Analytics

I designed a centralized patient data lake for a health insurance provider using AWS S3, Glue, and Athena. I ingested clinical, claims, and wearable data using AWS DMS, Kafka Connect, and Lambda functions.

I developed PySpark ETL pipelines on AWS Glue to cleanse and join data sets into a unified patient profile.

I used Delta Lake to manage schema evolution and versioning, ensuring traceability.

I also implemented de-identification logic using Python to adhere to HIPAA compliance.

To support analytics, I created Athena views and dashboards in QuickSight, allowing actuaries to analyze patient cohorts and claim patterns. The platform improved care personalization and reduced claim fraud.

Project 2: Claims Denial Analytics & Prediction Platform

I worked on building a data-driven solution to analyze and predict healthcare claims denials for a large insurance provider.

The core goal was to reduce the rate of denied claims, which cost millions annually.

I ingested EDI 837 claim files, enrollment data, and provider details into Azure Data Lake using Azure Data Factory and Kafka for near real-time integration.

I developed ETL pipelines in Azure Databricks using PySpark to clean, transform, and normalize data.

I built features such as CPT code combinations, diagnosis match rates, provider error history, and claim aging.

These were used to train an ML model in Azure Machine Learning Studio that could predict potential denials before submission 

The analytics layer was built on Power BI to help internal claim processors drill down by provider, location, or denial reason.

I also implemented Azure Monitor and Log Analytics for tracking ETL job health and data anomalies.

This platform led to a 27% drop in denial rates and improved processing efficiency by 35%.

Project 3: EHR Data Integration Using FHIR

I integrated multiple Electronic Health Record (EHR) systems using the FHIR standard.

I created data pipelines that ingested HL7 and JSON FHIR resources using Apache NiFi and Kafka.

I used Python scripts to validate schema, handle nested fields, and persist structured data into PostgreSQL and MongoDB depending on query patterns.

The data was further transformed and loaded into Redshift for analytics.

I enabled physicians to access consolidated patient histories through dashboards, reducing data retrieval time by 70%.

I also implemented audit logging and token-based access controls using OAuth2.0 to secure access.