Selected products, AI systems, research prototypes, and experiments — organized by what they are meant to show.
I build across practical AI systems, desktop tools, edge deployments, and experimental machine learning. This page highlights the projects that best represent those directions, then keeps smaller experiments in a compact archive.
The few projects that best represent my current direction.
A lightweight desktop visual companion and local-first animated overlay system.
A real-time visual agent where timing, data collection, and deployment behavior mattered as much as offline model metrics.
Local-first prompt version control with Git-style diffs, labels, metadata, and a SQLite backend.
An end-to-end TinyML audio classifier running on Raspberry Pi Pico and RP2040-class hardware.
Desktop software packaged and released for real users.
Desktop visual companion
A lightweight desktop visual companion and animated overlay experience. It started as a small alternative project and grew into a public release across GitHub, itch.io, and the Microsoft Store.
Creator screen recorder
A lightweight desktop screen recorder for creators producing vertical videos for TikTok, Instagram Reels, and YouTube Shorts. It focuses on fast capture-area selection, simple recording, and 9:16 export.
Local-first developer tool
A local-first prompt version control tool for tracking prompt changes, diffs, labels, and metadata in a portable SQLite file.
Research-oriented and system-level AI projects.
An RP2040 audio classification pipeline spanning feature extraction, training, and real-time firmware inference.
A real-time visual agent using a CNN with jump classification and press-duration regression heads.
A controlled comparison of continuous and discrete latent spaces, reconstruction quality, and failure modes.
A class-conditioned sketch generator with training diagnostics, checkpoints, and a small sampling GUI.
A temperature-trend detector combining rolling slope features, logistic regression, and embedded inference.
Useful side tools, prototypes, and focused technical builds.
Real-time 68-point facial landmarks with OpenCV DNN, Facemark LBF, smoothing, and CSV logging.
A GUI for recording examples, inspecting datasets, and debugging labels before model training.
A Streamlit tool for extracting dominant image colors with k-means, CIE LAB, and ΔE2000.
A local PDF assistant with chunking, similarity search, lightweight inference, and source attribution.
A Raspberry Pi Pico serial console for temperature telemetry, logging, LED control, and SCADA-style interaction.
An intentionally over-engineered parity experiment exploring wavelet features and unsupervised clustering.
Older learning projects, algorithm practice, and smaller experiments.
Five real-dataset regression exercises · Python · scikit-learn
Decision Tree, SVM, and KNN comparison · Metrics · scikit-learn
A playful character-level RL and reward-design experiment · Python
Algorithm and data-structure practice · Python · C++ · Bash
A minimal POSIX shell exploring parsing, fork, exec, and process waiting