Podcast
Practical LLM Engineering and RAG: Prompting, Evaluation and Real-World Workflows
Open original DataTalks.Club episode
Practical LLM Engineering and RAG: Prompting, Evaluation and Real-World Workflows
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 move from experimentation to reliable, production-ready LLM engineering and retrieval-augmented generation (RAG)? In this episode Hugo Bowne-Anderson — Head of Developer Relations at Outerbounds, longtime data scientist, educator, and host of Vanishing Gradients — walks through practical patterns for building, evaluating, and scaling real-world LLM workflows.
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 Kickoff & Hugo Bowne-Anderson Background
- 1:12 - Vanishing Gradients vs High Signal: Podcast Formats & Audiences
- 2:04 - From Academia to Industry: Biology Research, Python, and PyData
- 3:27 - Early Industry Work: DataCamp Curriculum and Product Roles
- 3:57 - Transition to Freelance: Consulting, Teaching, and DevRel
- 7:11 - Consulting vs Advisory: Hands-On Coding and Organizational Advice
- 8:24 - Driving AI Adoption: Loss Aversion and Dedicated Experimentation Time
- 9:28 - Everyday LLM Use Cases: Summaries, Translation, and CSV Workflows
- 11:11 - Prompting Best Practices: Role Prompts, Structured Output, and Timestamps
- 12:22 - Transcript Workflows: Gemini, Descript, Loom and Automation Tools
- 13:56 - Generator–Evaluator Pattern: Automated Quality Control for Outputs
- 17:38 - Scaling Transcript Pipelines: Automation with GitHub Actions
- 23:00 - Evaluation Sets for LLMs: Gold Tests, Size, Cost, and Representativeness
- 26:43 - Failure Analysis: Categorizing Errors and Prioritizing Retrieval Fixes
- 27:38 - Vibe Coding & Monitoring: Logging, Traces, and Debuggable MVPs
- 31:56 - Developer Tools & Assistants: GitHub Copilot, Cursor, and IDE Agents
- 33:14 - Embedded Agents in Workflows: Slack Integration and Proactive Assistants
- 40:12 - Agentic Value Beyond Chat: Actions, Documents, and Automation
- 44:26 - Prioritizing RAG: Quick Business Wins with Chunking and Embeddings
- 48:20 - Chunking Strategies: Fixed Length, Sliding Windows, and Context Rot
- 50:19 - When to Add Tooling: Moving from RAG to Agents and Tool Calls
- 53:34 - Practical Build: Email Assistant Example using Gmail API + RAG
- 56:21 - Four-Step Framework for Agents: Problem, Start Small, Data, Evaluation
- 57:41 - Memory Design: Retrieval-Based Memory vs Multi-Turn Conversation Memory
- 1:00:55 - Episode Wrap-Up: Key Takeaways, Courses, and Next Steps