MS Finance candidate at CU Boulder Leeds, CFA Level II candidate, and quantitative researcher on CU Quants' $500K student-run algorithmic trading fund. I build trading systems, financial data platforms, and analytical tools where precision actually matters.
A comprehensive evaluation of HMM applications to cryptocurrency markets, examining multiple training methodologies including the Viterbi algorithm for regime decoding, Forward-Backward (Baum-Welch) optimization, and online learning approaches designed to adapt to rapidly evolving market conditions. Develops both discrete and continuous emission sequence models, demonstrating that HMMs incorporating price, volume, and volatility measures across 3–7 hidden states achieve optimal predictive performance for identifying latent volatility regimes.
Financial data platform that extracts segment-level financials from S&P 500 SEC filings. Operating segments, geographic breakdowns, and peer comparisons through a dashboard, RESTful API, and Excel exports — at a fraction of Bloomberg's price.
6-layer stacked ensemble Hidden Markov Model for cryptocurrency regime detection and signal generation. Walk-forward validated with Optuna multi-objective hyperparameter optimization and parallel processing via joblib.
Production multi-asset trading system implementing ADX25_X2.5 strategy across 20 cryptocurrency assets. Desktop GUI with live execution, ensemble of 7 deep learning architectures including Transformer, TCN, LightGBM, Mamba, and xLSTM.
Sub-millisecond statistical arbitrage system on OKX perpetual futures. Numba JIT-compiled spread engine with Kalman filter dynamic hedge ratios and z-score mean reversion. 4-layer risk orchestrator with circuit breaker and real-time monitoring.
Desktop application for options pricing, Greeks computation, and risk analysis. Built with Tauri 2 (Rust backend) for near-native performance and React 19 frontend with interactive visualization of payoff diagrams and volatility surfaces.
Fine-tuning Google's MedGemma 1.5 4B vision-language model on dental radiograph datasets using 4-stage QLoRA curriculum training. Targeting 75%+ on MMOral-Bench, surpassing current SOTA (OralGPT at 66%). Trained on CU Boulder's Alpine A100 cluster.
Open-source scientific platform automating structural biology and analytical chemistry workflows. Protein conservation visualization, AI-powered NMR/mass spec annotation, 3D structure viewing, and data utilities.
Community forum for Yamaha XS750 triple motorcycle enthusiasts. Self-hosted platform for tech discussion, build logs, parts marketplace, and restoration resources.
Custom wire lead business for vintage motorcycle and automotive restoration. Build-to-order leads with selectable gauge, wire coating, color, and terminal ends — from quality Japanese bullet connectors to vintage plastic connectors. Expanding into multi-wire connectors, full custom harnesses, and LED conversion kits.
I'm an MS Finance candidate at CU Boulder's Leeds School of Business and a CFA Level II candidate. My work sits at the intersection of quantitative finance and applied computer science — building the tools, models, and systems that turn market data into decisions.
As a quantitative researcher on the CU Quants Fund — a real-money algorithmic market-making hedge fund at CU Boulder — I work on the Research team developing trading strategies across 9 crypto markets. The fund returned 39.5% annualized since its May 2025 inception, and I co-authored published research on Hidden Markov Models for cryptocurrency trading.
I also founded FinEdge Labs — a SaaS platform that extracts segment-level financials from SEC filings at institutional scale. I build systems that are production-grade: backtested, validated, and deployed.
When I'm not running models, I'm rebuilding a vintage Yamaha XS750 triple, producing electronic music, or somewhere in the mountains on skis.