Projects

Co-authoring a research publication on a Random Forest ML model that automatically detects geomagnetic events—enabling early warning systems for space weather that protect satellite infrastructure. Analyzing 10+ years of magnetometer data and validating hardware reliability for spaceflight environments.

• Developed Random Forest classifier achieving 80% recall for geomagnetic storm prediction using engineered features from SYM/H magnetometer data, outperforming traditional threshold-based methods.
• Engineered domain-specific features capturing pre-event quietness, compression amplitude, and ring current response to differentiate between storm commencements and impulses.
• Calibrate and maintain Helmholtz Coil apparatus for RM3100 sensor testing, validating hardware reliability for spaceflight environments.
• Implemented feature importance analysis revealing physical signatures as primary discriminators, demonstrating interpretable ML for space physics.

Collaborating with Curtis Ling (MaxLinear founder & CTO) to build a non-profit platform that connects student researchers to funding and enables donors direct access to ongoing university research.

• Co-designing the platform architecture using Python, React, and modern web technologies to serve the University of Michigan research community.
• Building systems that enable transparent, direct funding pipelines between donors and student-led research projects.

Engineering a secure, self-hosted financial platform that automates transaction classification, integrates real-time data via Plaid API, and provides dynamic visualization through React—reducing manual accounting workflows by 70%.

• Architected Python backend services that ingest and classify financial events with explainable outcomes, processing thousands of transactions with actionable insights.
• Integrated Plaid API for real-time financial data ingestion; PostgreSQL for persistent storage; React for dynamic, interactive visualization.
• Built workflow automations reducing manual accounting time by 70% and providing real-time visibility into spending patterns.