At Hero Housing Finance Limited, I designed and developed an automated ETL pipeline for financial datasets using Python, AWS S3, and Pandas. I processed business-critical datasets with validation and transformation workflows, including SCD Type 2 to preserve historical changes for audit-ready reporting.
I also integrated AI-based data quality checks and real-time Slack and email alerts to improve anomaly detection and reduce manual intervention.
On my cattle breed classification project, I trained a PyTorch model using ConvNeXt-Large to classify images across 16 Indian bovine breeds, reaching 79.39% validation accuracy. I also developed an unsupervised clustering system for researcher records, producing 2,500+ clusters and an automated evaluation and artifact-generation pipeline.
