Andrew Ng’s DeepLearning.AI, in collaboration with Chroma, an open-source embedding database company focused on AI-native solutions, has introduced a free one-hour course designed to teach participants advanced retrieval techniques specifically tailored for AI applications.
Called “Advanced Retrieval for AI with Chroma”, this short course is ideal for individuals with intermediate Python skills and a keen interest in mastering advanced retrieval techniques for extracting data from vector databases.
Led by Chroma cofounder Anton Troynikov, the primary focus is on refining information retrieval processes to ensure that the output from a database query is not only semantically similar but also highly relevant to the query and its intended application. Leveraging an LLM enhances the effectiveness of this traditional technique, and the course explores another form of expansion where the LLM suggests a potential answer to the query, subsequently included in the query itself.
Additionally, participants will also learn about cross-encoder reranking, a method to reorder retrieval results and prioritise those most relevant to the query, thereby improving overall results. Furthermore, the course covers the training and application of embedding adapters, introducing an adapter layer to reshape embeddings and improve elements pertinent to the specific application, leading to better retrieval outcomes.
Ng, who has democratised AI education for all through his free courses had earlier released courses on various topics like generative AI for all, LLM quality and security, vector database for LLMs and more. He has participated with several companies like Microsoft, Lamini, AWS, OpenAI for these training materials.
