Projects
These are some of the projects I've built, experiments I've tried, and/or tools that I keep returning to and tinkering with.
Latest work
Logic Gates Playground
Desktop sandbox for wiring basic logic gates into live circuits, complete with clocked inputs, truth-table export, and cross-platform build scripts (CMake + raylib).
Brainstorm — A Neural Chess Engine
Rust + libtorch engine that fuses deep network evaluation with alpha-beta search, hitting ~50k nodes/sec on an M1 and speaking UCI. Ships with a Python toolkit for self-play and PGN analysis.
NeuralNetPy
From-scratch deep learning playground powered by NumPy: dense and convolutional layers, modern activations, cross-entropy + MSE losses, and hand-rolled SGD/Adam optimizers.
Vision-Based Fight Detection
TensorFlow pipeline trained on the RWF-2000 dataset to flag aggressive motion in surveillance footage. Provides clip-level confidence, frame heatmaps, and a lightweight review dashboard.
PicoGrad.jl
Micrograd-sized autograd core in Julia with a lightweight MLP stack, gradient tracing visualizer, and notebooks that walk through every step of reverse-mode differentiation.