Advanced Retrieval for AI with Chroma

What You'll Learn

• Identify when queries are producing unsatisfactory results.
• Use a large language model (LLM) to enhance and refine query performance.
• Optimize and fine-tune embeddings using user feedback.

About This Course

Information Retrieval (IR) and Retrieval Augmented Generation (RAG) are most effective when relevant information is retrieved from the database in response to a query. This course covers advanced techniques for enhancing retrieval quality, especially for situations where semantically similar results fail to meet the query’s intent.

• Learn query expansion using an LLM to include related keywords and concepts, and incorporate suggested answers into the query.
• Improve retrieval results through cross-encoder reranking, prioritizing the most relevant content.
• Utilize embedding adapters to reshape embeddings and highlight application-relevant elements in retrieval results.

By the end of this course, you will understand advanced retrieval techniques, enabling you to implement precise and relevant query results in various applications.

Course Outline

• Introduction
  Introduction to retrieval techniques and the importance of precise query results.
• Overview of embeddings-based retrieval
  Introduction to embeddings and their use in information retrieval, with hands-on code examples.
• Pitfalls of retrieval - when simple vector search fails
  Examining the limitations of basic vector search and scenarios where it may yield suboptimal results.
• Query Expansion
  Techniques to expand user queries, including LLM-powered expansion for improved retrieval.
• Cross-encoder re-ranking
  Implementing cross-encoder re-ranking to prioritize the most relevant results in retrieval.
• Embedding adaptors
  Using embedding adapters to reshape and fine-tune embeddings, enhancing retrieval relevance.
• Other Techniques
  Overview of additional methods and best practices in advanced retrieval.

Who Should Join?

This course is designed for those with intermediate Python skills looking to enhance their retrieval capabilities within vector databases using advanced methods.