Deep Dive: Continuous Delivery for AI Applications with ECS
summary
Shows a working CI/CD pipeline that packages an MXNet inference app into Docker, pushes to Amazon ECR, and deploys via Amazon ECS with CloudFormation and CodePipeline; the live endpoint classifies an image as “suspension bridge” at ~62% probability. The repo and CloudFormation stack spin up networking, a build server and an ECS service in about 15–20 minutes. Each container is sized at 4 CPUs and 3 GB memory for inference, and the project is open sourced.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 2:05Where is deep learning today
- 3:40What about the application engineers
- 6:57What are the ingredients
- 11:12How do I make this repeatable and super fast
- 15:30Demo
- 17:27Deep Dive
- 19:05Code Pipeline
- 19:49Image Repository
- 21:57Behind the Service
- 23:26Demo Failure
- 24:43Demo Summary
- 29:19ECS is free
- 32:28Deep Learning
related talks
35mAccelerate deep learning in the cloud with custom ML environments
13mPremiere
12mLeveraging AWS and Kubernetes for Scalable AI/ML Workloads in the Cloud
42mCI/CD for Machine Learning
33mDeploying and Scaling Spark ML and Tensorflow AI Models
26mGet up and running with containers and Amazon ECS
34mFaster deep learning solutions...
12mServerless Deep Learning