RAG Agents in Prod: 10 Lessons We Learned — Douwe Kiela, creator of RAG
AI Engineer. Douwe Kiela led the original RAG paper at FAIR and now ships RAG into regulated enterprises. The talk is mostly about what stops working at scale — chunking strategies that don't survive 100k documents, "accuracy is table stakes, inaccuracy is the real problem," and why attribution and observability matter more than the embedding model. Good calibration before re-reading the article's eval and monitoring sections.
AI Expert note
Strong production cautionary material. Use it to pressure-test architecture decisions, especially permissions, attribution, observability and user trust, before debating embedding-model minutiae.
What you should get from this
Understand what breaks when RAG moves into regulated, high-stakes enterprise use.
Watch or know first
Know the basic RAG pipeline and why attribution matters.
Watch next
Continue through the same learning path with the next curated companion videos.
9 minutesRAG vs. Fine Tuning
Explain when retrieval is the right fix and when fine-tuning may actually help.
Watch next
8 minutesAnthropic's Claude Computer Use Is A Game Changer | YC Decoded
Decide where browser or computer-use agents might be commercially useful despite their operational risk.
Watch next
6 minutesAI Voice Agents: How They Actually Work & Why They Sound So Human
Recognize the core architecture of a voice agent and the failure points that affect customer trust in real calls.
Watch nextRelated videos
4 minutesIntroducing EmbeddingGemma: The Best-in-Class Open Model for On-Device Embeddings
32 minutesHow to Build Human-Centered AI Workflows in Localization with Shashi Bhushan
59 minutesFrom Hype to Habit: How Tech Companies Are Scaling AI Beyond the Experimental
41 minutesPrivate AI vs. Cloud: How Enterprise Leaders Can Make Smarter Build-or-Buy Decisions
Take it further
Hand-picked external courses that go deeper on this topic.
Coursera · DeepLearning.AIAI for Everyone
Six years after it launched, still the cleanest starting point for anyone who needs to understand AI without learning to code. No math, no jargon, no hype — you'll finish able to have an informed conversation about AI projects.
Coursera · The Wharton SchoolAI Strategy and Governance
Wharton's rigorous framing for executives making build-vs-buy decisions. Cuts through vendor pitches by focusing on the economics of AI deployment, algorithmic bias in hiring and operations, and the governance practices that survive an audit. Best taken before, not after, your next major AI procurement decision.