What is AI engineering and what do AI engineers even do?
Summary
Host Tejas Kumar defines AI engineering as building reliable, safe systems around probabilistic models, distinct from machine learning research. He addresses critical production challenges like hallucinations, cost spikes, and security risks, emphasizing grounded generation via RAG. The episode clarifies that AI engineers focus on behavior and integration rather than model training, making the role accessible to full-stack developers.
Topics discussed
Introduction and CodeCrafters sponsorship
The AI engineering crisis: bugs, hallucinations, and liability
Defining AI engineering vs. ML and full-stack roles
AI engineering as reliability engineering for behavior
Understanding and mitigating model hallucinations
Grounded generation and Retrieval Augmented Generation (RAG)
Real-time knowledge and tool-augmented truth
Context engineering: managing limits and safe compression
Agents, runtimes, and the Model Context Protocol (MCP)
Creating and structuring agent skills with skills.md
Multi-agent workflows and the OpenClaw case study
Conclusion: The future of trustworthy AI products
Listen ad-free on Castria