Podcast
MLOps at Scale: CI/CD, Reproducibility, Model Monitoring & Adoption Strategies
Open original DataTalks.Club episode
MLOps at Scale: CI/CD, Reproducibility, Model Monitoring & Adoption Strategies
Original Episode
Use these links for the canonical episode and media sources.
- Open the original DataTalks.Club podcast page
- Watch on YouTube
- Listen on Spotify
- Listen on Apple Podcasts
Episode Overview
How do you run MLOps at scale so models stay deployed, reproducible, and actually adopted? In this episode Raphaël Hoogvliets—who leads a 12-engineer team at Eneco and brings a career arc from agriculture into data science and MLOps—walks through practical approaches for CI/CD for ML, reproducibility, model monitoring, and adoption strategy.
People
Use these links to connect the episode to guest notes.
Chapter Summary
Use these checkpoints to decide whether to open the source transcript.
- 0:00 - Podcast Introduction
- 1:21 - Guest Overview: Raphaël Hoogvliets and Eneco role
- 2:34 - Career Path: From agriculture to data science and MLOps
- 8:41 - Agriculture technology, scale, and sustainability trade-offs
- 10:36 - Design Choices and Long-Term Tradeoffs in ML projects
- 13:37 - Speed vs. Robustness: trade-offs in MLOps delivery
- 14:05 - Team Coordination: why collaboration matters for ML at scale
- 16:58 - Key Team Roles: evangelists, tech translators, and technical leads
- 23:01 - Centralized MLOps as an enabling platform team
- 25:20 - Support Model: how MLOps assists product teams and ML engineers
- 27:56 - Adoption Strategy: iteration, feedback loops, and developer experience
- 32:46 - Building Trust: collecting pain points and delivering quick wins
- 36:55 - Measuring Value: KPIs, deployment frequency, and impact tracking
- 39:06 - Core Practices: CI, repo structure, parameterization, and testing
- 42:31 - Reproducibility: data versioning, traceability, and experiment capture
- 44:22 - Maturity Signals: when to introduce data versioning and governance
- 45:10 - Skill Mix: combining data science, SRE/devops, and platform engineering
- 48:41 - Getting Started: prioritize CI/CD and solve tangible pain points
- 51:21 - MLOps Toolset: experiment tracking, model registry, serving, and monitoring
- 53:08 - Dependency Management: package registries for reproducible deployments
- 56:50 - Container Strategy: Docker, Kubernetes, Databricks trade-offs
- 57:56 - Success & Failure Stories: deployment wins and integration freezes
- 1:00:54 - MLOps Defined: operationalizing machine learning in business
- 1:01:58 - Core Challenge: keeping models deployed, monitored, and maintained
- 1:02:42 - Closing Remarks and next steps