client-side-ml

Chrome extension that classifies Sentry errors with in-browser ML. Trains locally, learns from your corrections, zero backend.

SVG-native local AI companion powered by browser-side Transformers.js and ONNX. No Python backend, no hosted inference.

A privacy-first, fully offline neural machine translation application supporting 200+ languages. Built on Meta's NLLB-200 model, this tool runs entirely in your browser without requiring server infrastructure or internet connectivity after initial model download.

Run small LLMs directly in your web browser, no cloud computing needed.

Browser-based clothing classifier using a custom Teachable Machine model with entropy-based confidence filtering (the foundation for PICKyourFIT)

An interactive, browser-based Real-Time Image Classifier using TensorFlow.js. Train a custom neural network entirely client-side using your webcam, complete with live predictions, model exporting.

Real-time emotion detection and facial expression analysis web app powered by MediaPipe and ONNX Runtime. Features live webcam tracking and interview coach mode.

My portfolio and engineering journal. A statically generated Next.js site with an MDX writing section and a Lab of machine-learning experiments that run entirely in the visitor's browser, with no inference server behind them.

AI-powered personal finance dashboard, Auto-categorizes expenses, predicts spending trends, and provides intelligent insights.

Interactive UX concept exploring a portable pocket terminal in the browser using Svelte, TypeScript & Transformers.js. Features client-side AI chat (Qwen2.5-Coder via Web Workers/WebGPU), Conway's Game of Life with entropy metrics, live Bitcoin ASCII charting, and system diagnostics.