Building Machine Learning Microservices & MLOps using Union ML
summary
UnionML claims to turn notebooks into microservices with decorator functions (reader, trainer, predictor, evaluator) so data pipelines avoid boilerplate. It is built on Flight, installs with pip install unionml, and serves via unionml server (demo runs on localhost:8000). Demos use MNIST with logistic regression and an Iris KNN hitting ~96% accuracy; integrations include FastAPI, AWS Lambda and BentoML, supports TensorFlow/PyTorch/sklearn/ONNX, can request GPUs, and the CLI initializes ready apps.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:02ML lifecycle overview
- 3:46MLOps vs DevOps and model challenges
- 8:04Union ML goals and standardization
- 10:51Flyte and workflow automation
- 14:23Building ML microservices
- 18:56Union ML deployment advantages
- 24:18Live demos
- 29:31Q&A
related talks
08mTour of UnionML: An Open Source framework for Building Machine Learn Microservices
53mClearML: Your R&D on MLOps! From zero to hero in two lines of code
41mMLOps in 30 Minutes
33mBuilding Resilient (ML) Pipelines for MLOps
32mSimplify End-To-End MLOps with PostgresML
41mBuilding Serverless Machine Learning Models in the Cloud
47mMachine Learning in Production: An Intro to MLOps.
25mTinyML as-a-Service