The following are production-grade projects and startups I’ve developed outside my career in quantitative finance (which remains under NDA). They include a natural language processing platform, a commercial-grade sound visualization app, and a real-time AI assistant deployed at a working spa business.

Real-Time Conversational AI with LLMs and External Knowledge Integration

Thy Spa AI (external link, 2025) — A real-time voice Q&A assistant built using GPT-4 Mini Realtime, designed to provide expert advice in the health & wellness domain. This project explores the integration of LLMs with external knowledge and real-time streaming input. It’s a conversational agent built to feel helpful, fast, and natural — combining applied AI with user-centered design.

GPU-Accelerated Audio Visualization with Custom Signal Processing

VSound App (2018-now) — A sound visualization app available on macOS desktop and iOS, built using C++, Swift, and Metal for real-time GPU-accelerated rendering. Based on a custom frequency detection algorithm, VSound combines signal processing with creative coding to produce responsive, elegant audio visuals. This project explores both performance optimization and cross-platform design.

Probabilistic Topic Modeling with LDA and Bayesian Inference

Topic Models (2011-2014) — A series of experiments in natural language processing and unsupervised learning, focusing on topic extraction from large text corpora using probabilistic models (such as Latent Dirichlet Allocation). Originally used to analyze 10-K corporate filings, with broader implications for document clustering and knowledge discovery.