Applied Machine Learning in Elixir
summary
Claims that going custom beats big-name agents and demonstrates an Elixir ML stack: NX for tensors and autodiff, Axon for neural nets and training, Bumblebee for Hugging Face models, and Livebook plus Phoenix LiveView for deployment. Shows specific backends (EXLA, Torchex, MLX), compiles to CPU/GPU/wasm, and trains a tiny XOR network, serves it with batching and returns correct thresholded outputs. Debugging hooks and a GitHub repo are provided and the demo completes successfully.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:42NX overview
- 2:48ML libraries in Elixir
- 5:23Numerical computation in NX
- 8:41LiveBook setup
- 10:06Defafn basics
- 13:42Gradients in Defend
- 15:29Neural network training
- 18:45Model inference
- 20:35Serving models
- 21:38LiveView demo
- 22:41Q&A
related talks
47mLivebook & Elixir: Where AI, Web & Concurrency Meet
39mIntroducing Elixir: Your entire web stack
28mMachine Learning Live
54mMachine Learning for Developers
45mKeynote: Panel(...)
20mScaling Edge AI with Elixir and Nerves
42mMachine Learning with Elixir and Phoenix
10mML-based Log Analysis for Faster Debugging