Data Brew by Databricks Retrieval, rerankers, and RAG tips and tricks | Data Brew | Episode 39
10 snips
Feb 20, 2025 Andrew Drozdov, a research scientist at Databricks specializing in Retrieval Augmented Generation (RAG), dives deep into enhancing AI models. He discusses overcoming LLM limitations by integrating relevant external information and optimizing document chunking and query generation. The conversation also highlights the significance of embeddings and fine-tuning techniques for retrieval systems. Additionally, Andrew shares insights on improving search results with re-ranking strategies and the application of RAG methods in enterprise AI for better domain-specific outcomes.
AI Snips
Chapters
Transcript
Episode notes
Better Query Generation
- Improve RAG pipelines by generating better queries.
- Use the language model itself to rewrite user queries into more concise keyword-based searches.
Personalization in RAG
- Personalization can occur at both the generation and retrieval stages in RAG.
- This allows tuning both the response style and the type of document retrieved, like Wikipedia versus internal docs.
Recommendation System Challenges
- At eBay, recommendations based on purchase history proved challenging due to diverse items.
- Balancing weights for previous purchases was key, as buying a PlayStation doesn't necessitate another PlayStation recommendation.