Publishing a Machine Learning model as an API product
summary
Metamaze exposes OCR/ML as an API with an orchestrator that launches Kubernetes training jobs and puts versions into an Azure blob model repository; deployments scale by config and a node-based orchestrator routes calls to tenant-specific models. Automation targets 90% automation initially (some projects start at 50% and can approach 99%), uses an 80% confidence cutoff to fall back to human-in-loop correction, and pairs RPA for customers whose IT teams take months or years to integrate.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:43What is Machine Learning
- 3:03Limitations of Machine Learning
- 3:44Who are we
- 4:13Machine Learning Model
- 5:54Current Process
- 6:51Humans
related talks
42mImproving Machine Learning Development Reliability
26mManaging ML Models @ Scale
43mA lap around Azure Machine Learning services
13mApplying Azure Machine Learning service to drive success of API programs
41mLeveraging Automated Machine Learning to enable anyone to develop Machine Learning solutions
32mAutomating Production Production Level Machine Learning on AWS
32mProductionize Machine Learning on Azure
59mAlgorithmia in Action: Take your ML models from training to production