A CI/CD Framework for Production Machine Learning at Massive Scale (using Jenkins X and Seldon Core)
summary
Shows a hands‑on pipeline that trains a text‑classification model, packages it with Seldon Core into a Docker microservice, and promotes versions via Jenkins X + Tekton using GitOps PR commands. Demonstrates canary and shadow deployments, CI via GitHub PR comments, and model monitoring with Alibi for explainability, outlier and drift detection. Describes scaling to thousands of heterogeneous models, cites Seldon’s ~2 million installations, and shows the pipeline running end‑to‑end with reproducible provenance and rollbacks.
AI-written summary · about
chapters (25)
- 0:00Intro
- 0:11Hello, my name is Alejandro
- 0:52Why is a prod ML so hard?
- 1:47Challenges at scale
- 2:34End-to-end Approach
- 3:23Principles for responsible AI
- 3:51What is MLOps?
- 4:52Key Requirements
- 5:56Introducing Seldon Core
- 7:10Auditability and Provenance
- 7:42Unified Model Metadata
- 8:04Metrics Monitoring
- 8:24Audit Trails on Data
- 8:43Continuous Delivery?
- 9:14Introducing Jenkins X
- 13:36Jenkins X MLOps CD
- 15:48Hands on implementation
- 16:08Compliance at Scale?
- 16:36Introducing Alibi
- 17:16Inference Data Assessment
- 17:26Model Assessment
- 17:48Production Monitoring
- 18:15Adversarial Security
- 18:43Explainability as Infrastructures
- 19:40Outlier Detection as Infrastructure
related talks
21mMastering the DevOps Machine Learning Pipeline
34mServing ML Models at Scale with Seldon and Kubeflow
18mBeyond Inference: Bringing ML into Production
33mDeploying and Scaling Spark ML and Tensorflow AI Models
33mCompliant Data Management and Machine Learning on Kubernetes
42mCI/CD for Machine Learning
57mDevOps for Machine Learning
24mFrom Model to Microservice - Machine Learning at Scale