MLDS 2019: Madhusudan Shekar of Amazon at Machine Learning Developers Summit 2019
summary
AI reduces the cost of making predictions, not intelligence, and the concrete AWS stack aims to democratize AI by shipping managed pieces you can use today: SageMaker for end-to-end training and deployment, Ground Truth for scalable human labeling, and DeepRacer plus RoboMaker for reinforcement learning practice. Claims include optimized TensorFlow cutting ResNet‑50 training from ~30 to ~14 minutes and a p3.3d instance delivering ~1 petaflop for about $24/hour; elastic inference and Neo shrink inference cost and model size for edge deployment.
AI-written summary · about
chapters (11)
- 0:00Start
- 1:26What Does Ai Make Cheaper
- 8:53Scalability
- 11:51Ai Services
- 14:18Forecasting and Recommendations
- 15:16Forecasting
- 17:59Why You Should Not Build Machine Learning on Your Laptops
- 21:10Data Leakage Prevention
- 22:50Scaling Tensorflow
- 26:11Scaling Machine Learning
- 33:19Ss Deep Racer
related talks
19mUsing Amazon SageMaker to develop and deploy ML models
44mPragmatic Machine Learning in the Cloud
42mFrom Zero to Machine Learning in 45 Minutes
06mAI and machine learning at Amazon
40mCloud-based Machine Learning Offerings - Beginner's Perspective
54mDeep learning on AWS with NVIDIA: From training to deployment
20mMArk: Exploiting Cloud Services for Cost-Effective, SLO-Aware Machine Learning Inference Serving
33mBuild, train, and deploy ML models using Amazon SageMaker