Beyond Inference: Bringing ML into Production
summary
Argues that model value starts after training and demonstrates a runnable pipeline: a containerized image classifier plus an anchor explainer deployed as two pods under one Seldon Core Deployment on OpenShift/Open Data Hub, wired to JupyterHub for iteration and to Prometheus + Grafana for metrics. YAML glue and REST gateway take months to tame, but the demo returns an 84% “siamese cat” prediction with 2% “paper towel” risk and exposes drift, A/B routing and ensembling capabilities, so the setup mostly works despite some clunky edges.
AI-written summary · about
chapters (5)
related talks
21mMastering the DevOps Machine Learning Pipeline
22mFrom experiment to production - a journey of a machine learning model
33mCompliant Data Management and Machine Learning on Kubernetes
01h 21mBuilding a ML Model Observability pipeline
25mContinuous ML Applications in Production
32mA CI/CD Framework for Production Machine Learning at Massive Scale (using Jenkins X and Seldon Core)
29mDemystifying Machine Learning in Production: Reasoning about a Large-Scale ML Platform
29mMachine learning in production