Vamshi KrishnaPEDDAGOLLA

Senior Data Engineer, Programming.com

Bengaluru, IndiaJuly 2025 – Present

• Built pipelines for a personal loans lender covering disbursement, repayment tracking, NPA monitoring and collections, on AWS Glue, S3, Lambda and Snowflake.
• Converted legacy Python ETL scripts to PySpark on AWS Glue for distributed processing of large banking datasets.
• Designed Snowflake warehouses with clustering, partitioning and role-based access control to protect sensitive financial data.
• Wrote the allocation logic that assigns collection priorities and distributes accounts to loan officers by branch, pincode and priority.
• Automated alerts to risk officers and field teams, and standardised master data across systems.

New Jersey, USAFebruary 2024 – February 2025

• Built Azure Data Factory and Databricks pipelines on Delta Lake (bronze, silver, gold) delivering PHI-compliant datasets from EHR, LIS and HL7/FHIR sources.
• Cut nightly PySpark ETL runtimes by 30% across 2TB+ of clinical data and automated releases with Azure DevOps.
• Modelled healthcare data marts and built Power BI dashboards with row-level security for care teams and leadership.
• Added Great Expectations checks for HIPAA audit readiness and real-time alerts with Event Hubs.
• Trained readmission, sepsis and length-of-stay models with scikit-learn and XGBoost.

Project Engineer, Wipro Technology

IndiaFebruary 2022 – October 2023

• Ran real-time ingestion on Amazon Kinesis (5 GB of events a day) and PySpark Glue jobs loading S3 and Redshift.
• Designed star and snowflake schemas with Type 1 and Type 2 slowly changing dimensions in Redshift.
• Added data-quality rules, QuickSight dashboards and CloudWatch monitoring, with IAM least-privilege access and KMS encryption.

Patient disease risk prediction

Python, scikit-learn, Flask, Docker, AWS Fargate, MLflow

• Prepared the Diabetes 130-US Hospitals dataset and compared logistic regression, random forest and gradient boosting with stratified 5-fold cross-validation. The tuned model reached 87% accuracy.
• Served the best model through a Flask API in Docker on AWS Fargate, with MLflow tracking experiments and versions.

Product segmentation pipeline

TensorFlow, Mask R-CNN, UNet, CVAT

• Raised retail image-classification accuracy by 30% at under 40 ms per frame with a segmentation stack.
• Annotated 1,000+ product images with a CVAT workflow that cut labelling effort per image by 60%.