in-browser-ml

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

Browser-side RAG demo — runs embeddings entirely in the browser via transformers.js (@xenova), no backend. React + Vite + TypeScript.

PyTorch-shaped machine learning that runs in a browser tab — hand-written WGSL on WebGPU, a numpy core, and a Python binding, all held to the same values frozen from real PyTorch.

Bilingual (English + Bangla) sentiment analysis platform that runs entirely in your browser. Features an explainable rule-based engine, real in-browser ML training (LR / NB / SVM) with confusion matrix and per-class F1, live visualizations, and an optional hardened LLM chat. Offline-first, built with React + TypeScript.

Do you need a transformer to forecast? Five tiny models (none over 80k params) vs published transformer results on the ETTh1 long-horizon benchmark — with an in-browser demo.

An AI that teaches itself to drive — a PPO-trained Transformer running live in your browser (WebGPU), where you can also wipe the brain and watch a new network learn from scratch. No hand-coded rules.

Upload any CSV, get a trained + evaluated + explained model. Real scikit-learn running in your browser via Pyodide — no backend.

Teach a face-expression classifier your own face in seconds. Softmax regression with Adam, written from scratch, trained in the browser on MediaPipe blendshapes. Works without a camera too.

Multi-label emotion classification on GoEmotions: a 3.8 MB linear model at 96% of BERT, served from FastAPI and in-browser JS

Draw a digit and a 109k-parameter neural net reads it — hand-written forward pass in plain JS, int8 weights, 97.58% on held-out MNIST. No TensorFlow.

Browser palm-crease segmentation: a 5.5M-param ONNX UNet that traces heart/head/life lines client-side. Live demo + full training pipeline.

Real-time, in-browser facial expression reader with calibrated confidence, literature-grounded cognitive states, and FACS-grounded AI explanations.

Live in-browser ASL alphabet classifier (Next.js + ONNX Runtime Web) with honest cross-dataset accuracy reporting

Email-powered job application tracker — syncs Gmail & iCloud, classifies emails with ML, tracks your job search pipeline