Challenges of machine learning: The new ML lifecycle
summary
Claims that only 10% deployed of organizations and that 50% never see the light frame the scale of the problem, with lab-to-prod delays often hitting 18 months. Notes concrete pain points: heterogeneous CPU/GPU training, Kubernetes and CI/CD mismatches, throw-over-the-wall handoffs, hidden technical debt. Recommends planning deployment and SLAs from day one, using registries, versioning, audit logs and pipelines, and offers an Algorithmia model deployment platform that can run inside enterprise networks for governance and reuse.
AI-written summary · about
chapters (13)
- 0:00Introduction and speaker background
- 1:27Production ML challenges and lifecycle
- 3:53Stuck in the lab problem
- 6:18Economic challenges in ML deployment
- 8:23Wrong incentives and goals
- 10:52Wrong teams and skill gaps
- 13:36Lack of proper champions
- 14:39Wrong technology stack
- 15:41Technical architecture for ML
- 18:30Operationalizing ML in enterprises
- 20:14Common pitfalls and solutions
- 20:56Algorithmia platform overview
- 21:43Q&A session
related talks
52mThe ML Lifecycle in Combat Conditions
17mContinuous deployment for ML: The new software development lifecycle
41mChallenges in incorporating ML models
29mDemystifying Machine Learning in Production: Reasoning about a Large-Scale ML Platform
30mThe State of Production Machine Learning in 2023
27mKeys to Building Machine Learning Systems
21mHow The Application Of Machine Learning Is Changing The Way Companies Use Data
24mChaos and Pain in Machine Learning, and the ‘DevOps for ML Manifesto'