Level Up Your Machine Learning Lifecycle
summary
Claim: reuse a unified image/object pipeline to turn a pet-breed classifier into a license-plate detector, cutting development time and reducing need for “unicorn” hires. Implementation uses Dockerized environments, TensorFlow/PyTorch/scikit-learn models, AWS S3/GCS handlers, pytest for unit tests, and an engineer-facing one-button configuration plus training/eval APIs. Specifics: region-proposal + cropping/resizing reuse, 80/20 train/test split, segmentation excluded as a separate product. Product portfolio exists and is presented as working.
AI-written summary · about
chapters (6)
- 0:00Start
- 5:10Machine Learning Life Cycle
- 5:42Three Stages
- 7:46The Machine Learning Life Cycle to the Next Level
- 15:34Region Proposal
- 26:00Type of Model Output
related talks
13mMLOps Pipelines to Deliver Tf.Lite Models to Distributed Compute Pools at the Edge
42mDeep learning beyond the learning
46mML Use Cases at Think Big Analytics
40mHow to Implement a Production Ready Deep Learning Model in 12 Weeks
40mServerless Machine Learning 101
46mAccelerating Machine Learning DevOps with Kubeflow
30mRealizing End to End Reproducible Machine Learning on Kubernetes
23mAutomation: Build A Training Pipeline