Podcast
Building Data Products at Scale: Intake, A/B Testing, and MLOps in a Marketing Organization
Open original DataTalks.Club episode
Building Data Products at Scale: Intake, A/B Testing, and MLOps in a Marketing Organization
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 prioritize data product work, validate models in production, and keep them monitored without overwhelming stakeholders? In this episode, Ioannis Mesionis, Lead Data Scientist at easyJet and head of their MLOps efforts, walks through a practical data product operating model for tackling those challenges.
People
Use these links to connect the episode to guest notes.
Chapter Summary
Use these checkpoints to decide whether to open the source transcript.
- 1:40 - Episode introduction & guest Ioannis Mesionis (EasyJet lead data scientist)
- 2:34 - Career origin & early projects (mathematics degree, master’s, internship model)
- 7:23 - Lead Data Scientist role: partnering with Digital Customer & Marketing
- 8:32 - Stakeholder collaboration: weekly embedded meetings and observation
- 11:15 - Business domain knowledge: PPC, SEO, keywords and conversion optimization
- 14:00 - Operating model for data products: four-phase funnel and accountability
- 15:23 - Project intake & prioritization: “single front door” and cross-functional
- 17:37 - Definition of Done: template, KPIs, success criteria and fail‑fast checks
- 20:54 - Inception & EDA: data access, GDPR considerations and feasibility assessment
- 21:12 - Data science vs analytics: choosing technical approach and leads
- 22:48 - Research & development: modeling work, sprint planning and Kanban usage
- 25:17 - Pilot & A/B testing: validating models against baseline KPIs and feedback
- 27:25 - Production rollout: spectrum of production and evolving MLOps capabilities
- 28:18 - Organizational structure: domain-focused lead data scientists (scheduling,
- 30:21 - Handling uncertainty in ML: MVPs, estimation practices and Kanban preference
- 35:38 - Sprint cadence: planning, stand-ups, bi‑weekly demos and stakeholder demos
- 38:17 - Estimation techniques: T-shirt sizing, Planning Poker and Fibonacci points
- 40:49 - Stakeholder engagement strategy: invite to demos, not daily stand-ups
- 41:33 - Communicating technical results: simplifying concepts for non‑technical audiences
- 45:10 - Developing soft skills: practice, analogies, feedback and ChatGPT as a helper
- 48:38 - MLOps Zoomcamp takeaways: motivation for hands‑on MLOps learning
- 49:10 - MLOps tooling overview: MLflow, Prefect, Airflow and engineering exposure
- 53:33 - Model monitoring with Evidently: drift detection and integration plans
- 55:11 - Monitoring dashboards & alerts: Tableau quick solutions and custom emails
- 57:09 - Recommended resources: Cassie Kozyrkov (Decision Intelligence) and textbooks