Neil Gandhi

Bachelor's

Computer Science

UPS

ML/Automation Developer Intern

• Developed “Trace”, a production-grade fraud detection system for uncovering organized fraud rings, achieving 90%+ detection accuracy. System demonstrates potential to reduce annual fraud loss by $5.5M across 1M+ UPS users. • Engineered novel temporal and network-based features to detect account takeover and credit card fraud; addressed extreme class imbalance to boost minority-class recall. • Validated performance using 5-fold cross-validation (Shuffle AUC: 0.456), ensuring robustness across imbalanced datasets. • Implemented using PyTorch Geometric, HGTConv layers, and heterogeneous multi-relational graphs for graph learning. • Automated the Solution Group intake workflow using Azure Functions, Power Automate, and Python, eliminating 60+ hours of manual effort biweekly and standardizing solution onboarding across teams. • Built an automated pipeline to extract data from Word documents, populate SharePoint lists, and auto-generate executive-ready presentations.