Building Machine Learning-powered Applications with JS
summary
Claims that production ML apps run end-to-end in JavaScript, showing browser and on-device inference with TensorFlow.js and ONNX Runtime for Web, and cloud training/deployment via AWS SageMaker JavaScript SDK and Lambda. Trains a classifier on ~100,000 images reaching ~90% accuracy, warns about model theft and large model sizes for client delivery, and demonstrates using parallel cloud jobs to cut 100 sequential experiments (100 hours) down to about 1–1.5 hours using concurrent instances.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:22Speaker background
- 2:03How ML works
- 5:39Why JavaScript for ML
- 8:29ML inference in browser
- 13:49ML with JS frameworks
- 16:21Deploy ML in mobile apps
- 18:53Cloud advantages for ML
- 22:51AWS SageMaker with JS SDK
- 25:00Serverless JS deployments
- 27:26Choosing ML tools in JS
- 30:18Train ML: server vs browser
- 33:06Closing remarks and key takeaways
related talks
15mThe future of machine learning using JavaScript
23mThe new era of ML in JavaScript
39mMachine Learning Magic for Your JavaScript Application
18mMachine Learning with Javascript
23mNode.js = Machine Learning
24mThe Future of Machine Learning & Javascript
19mML at Client Side
32mUsing Machine Learning Models in JavaScript Application