Work Experience

Built GT-Suite model for 4-cylinder diesel engine. Achieved brake torque error within ±3%, model accuracy >97%. Learned that technical precision matters, but understanding the system matters more.

Stakeholder research revealed miners needed hands-free safety communication. Pivoted from inner-braces to glove-based haptic feedback after field study. First lesson in product-market fit.

Designed impact-absorbing phone case using auxetic structures. Learned CAD, prototyping, and that good design requires understanding material properties as well.

Cleaned 3,300+ messy alumni records. Built Random Forest models for career prediction. This project opened the analytics path.

Built ML framework to quantify aesthetic principles using YOLOv8. Study with 101 participants revealed simplicity and contrast drive preference. Proved you CAN quantify aesthetics, contrary to what everyone said.

Led market research for autonomous underwater vehicle. Researched 22 competitors, interviewed potential users, won 4 government competitions. Learned ownership the hard way in messy startup environment.

Applied Nielsen's 10 heuristics across 5 core screens. Identified 47 usability issues with actionable recommendations. Learned to critique design systematically, not just say "it feels off."

Analyzed India's semiconductor opportunity. Competitive analysis of 22 global players, market sizing, multi-phase roadmap. Learned to think about markets, not just products.

Observed print shop operations to find bottlenecks operators couldn't articulate. Proposed automation based on actual workflow, not assumptions. First real ethnographic research project.

Analyzing end-to-end AML examination workflows spanning 100+ annual regulatory reviews. Scoping AI-assisted evidence aggregation patterns for compliance tools with focus on traceability and examiner trust.

Built neural network for loan default prediction achieving 88.8% AUC-ROC with interpretable counterfactual explanations. Generated actionable recommendations showing applicants exactly what changes would flip their outcome.

Built AI tool that transforms unstructured research into evidence-based product decisions. Conducted user research with 4 PMs revealing systematic bias. System extracts insights, searches for counter-evidence, and generates transparent reasoning.

Built 2-layer neural network on Census data. Achieved 85.6% accuracy using ensemble of 5 k-fold models with optimized thresholds for class imbalance. First serious deep learning project.

Built collaborative filtering system for vacation property matching. Learned recommendation algorithms, user preference analysis, and how to handle sparse data.

Regression, hypothesis testing, A/B testing, experimental design. Learning when data actually supports a decision and when it doesn't. Most important lesson: knowing what you can't conclude.

Built to tell my story authentically.

Looking for opportunities that value curiosity, systems thinking, and the ability to bridge technical and human problems. Probably in finance or tech.