The State of AI and Machine Learning Operations in 2019
summary
A 2019 survey of MLOps that names concrete tools and shows working examples: reproducible pipelines using DVC, Pachyderm, Seldon with persistent volumes for full provenance; an NLP pipeline that trains on Reddit hate‑speech and streams Brexit tweets; a PyTorch Hub example that containerizes and deploys models to Kubernetes with about ten lines of config; and an explainer pattern using Alibi anchor explanations to audit an income classifier (>50k) running in production. Metrics, logging and GitOps integrations are demonstrated end to end.
AI-written summary · about
chapters (10)
- 0:00ML Ops introduction
- 1:03Machine learning operations rise
- 3:52ML engineer role
- 5:20Reproducibility in ML
- 6:22NLP pipeline case study
- 12:02Model orchestration with Kubernetes
- 16:57Deploying models with Seldon and PyTorch Hub
- 25:13Explainability importance
- 28:20Explainable model patterns
- 34:36Closing and ecosystem
related talks
34mThe State of AI and Machine Learning Operations in 2019
53mExperience with an end-to-end MLOps pipeline: Practical insights and 3D architectural perspectives
13mMLOps: From AI Research to Production-Ready Systems
44mDevOps for Machine Learning - why is it different?
13mMLOps at Scale
41mTHE STATE OF MLOPS - MACHINE LEARNING IN PRODUCTION AT ENTERPRISE SCALE
26mAre you ready for MLOps? 🫵
26mKubernetes and MLOps for Scalable and Reproducible Generative AI