Podcast
Build Explainable and Actionable AI/ML Systems: Industrial PhD, Trust Theory & Production Deployment
Open original DataTalks.Club episode
Build Explainable and Actionable AI/ML Systems: Industrial PhD, Trust Theory & Production Deployment
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 build ML systems that business teams trust and can act on? In this episode, Polina Mosolova — a data scientist at SAP who completed an industrial PhD building end-to-end ML pipelines — demonstrates how to bridge research and production through explainable AI grounded in organizational trust theory. Drawing from her churn prediction research, Polina shows how the ABI framework (Ability, Benevolence, Integrity) transforms model explanations into actionable business interventions.
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 - Episode Introduction & Overview
- 1:14 - Guest Introduction: Polina Mosolova — Industrial PhD and Churn Prediction
- 2:05 - Career Journey: Industrial PhD to Full-Stack Data Scientist at SAP
- 7:19 - Role Evolution: From Full-Stack Data Scientist to MLOps Specialization
- 9:19 - PhD Practice: Building End-to-End ML Pipelines During Doctoral Research
- 10:34 - Dual Goals: Balancing Academic Research and Production Deliverables
- 12:33 - Dissertation Focus: Churn Prediction Informed by Organizational Trust Theory
- 14:02 - Production Challenges: Deploying Research Models in Industry
- 17:57 - Supervision & Stakeholders: Academic and Company Support Structures
- 19:05 - Research-Industry Bridge: Academic Conferences and Summer Schools
- 20:37 - Time Management: Balancing PhD Writing with Industrial Responsibilities
- 24:38 - Finding Industrial PhDs: Prevalence, Companies, and How to Search
- 27:41 - Practical Tips: Job Postings, Language Requirements, and Application Search
- 29:52 - Organizational Trust Theory: ABI Framework — Ability, Benevolence, Integrity
- 34:36 - Pricing, Contracts, and Trust Dynamics in Subscription Services
- 38:19 - Linking Organizational Trust to Explainable AI and Feature Design
- 41:54 - Actionability: Turning Explanations into Usable Business Interventions
- 44:03 - Definitions: Interpretability vs Explainability vs Actionable ML
- 47:22 - Model Choices: Glass-Box Models, Generalized Additive Models, Neural Additive
- 49:00 - Explainability Tools: Random Forest + SHAP — Explainable vs Interpretable
- 50:47 - Computer Vision Explainability: Activation Maps and Human Interpretability
- 51:48 - Summary: Interpretable Models, Explainable Outputs, and Actionable Decisions
- 52:39 - Audience Matters: Explainable Feature Spaces and Tailoring Explanations
- 55:23 - Explainable AI and Trust: User Confidence, Provenance, and Transparency
- 57:43 - LLMs and Hallucinations: Explainability Challenges Versus Tabular Models
- 58:18 - Measuring Trust: KPIs, Proxies, and Ethical Constraints
- 1:00:29 - Business Relevance: Practical Proxies for Trust and Prioritizing Product