Machine Learning in the Browser: Fast Iteration with ONNX & WebAssembly
summary
Claims that shipping edge demos fixes the integration gap where up to 80% of models never reach product; shows concrete pipeline: train in Python, export via scikit‑learn->ONNX, run with ONNX runtime web compiled to WebAssembly (optionally WebGPU) or use transformer.js for LLMs. Demonstrates a codegen tool (model ship / uvx) that produces a static web app, deploys to GitHub Pages, and runs locally with no outbound requests, proving the approach works end to end.
AI-written summary · about
chapters (14)
- 0:00ML in the browser
- 2:53Deploying ML models
- 3:35ONNX & WebAssembly
- 4:40ONNX benefits
- 6:26Converting to ONNX
- 7:53Visualizing ONNX models
- 8:38ONNX Runtime & WASM
- 9:00WebAssembly fundamentals
- 12:32Live demo ONNX Runtime Web
- 16:10ONNX in web apps
- 18:46LLMs with Transformers.js
- 20:18Model Ship automation
- 23:15GitHub Pages deployment
- 24:45Key takeaways
related talks
29mML/AI models in JavaScript environment
51mMaking neural networks run in browser with ONNX
45mPrivacy-First In-Browser Generative AI Web Apps
36mApplied AI on the edge
41mEscaping the Cloud: High-Performance AI in your Browser
33mBringing AI to the Heterogeneous Edge with WebAssembly & ONNX
39mWebAssembly-Powered Machine Learning
32mEmpower WebAssembly ML Over Diverse AI Hardware Accelerators Outside the Browser