Скачать или смотреть RAG Antipatterns in the wild and how to fix them

In this talk, Skylar Payne, Founder of Wicked Data LLC, goes into common RAG anti-patterns that he’s encountered across his consulting work. From data collection issues to generation problems, Skylar shares real-world examples from medical chatbots, financial news summarization, academic research assistants, and more, revealing the most common mistakes that cause AI projects to fail and how to fix them.

Skylar has personally debugged retrieval systems across every scale, and brings battle-tested patterns that bridge the gap between ML experiments and reliable products.

Wicked Data LLC specializes in guiding companies with in-house development teams to collaboratively establish and execute a confident data and AI strategy that drives real business outcomes.

If you want to learn more about improving rag applications check out: https://improvingrag.com/

TIME STAMPS

00:00 Introduction
02:56 Consulting Case Studies Overview
08:01 Anti-Patterns in AI Projects
10:15 Data Collection and Curation Issues
14:23 Extraction and Enrichment Challenges
20:21 Indexing and Storage Problems
23:59 Retrieval and Query Handling
27:23 Understanding Task Breakdown in AI Systems
30:02 Evaluating for False Negatives
35:54 Avoiding Face Palm Results in Retrieval
41:49 Addressing Hallucination in Outputs
43:28 Q&A Session and Tool Preferences
49:05 Final Thoughts and Future Discussions