{"podcast":{"title":"MLOps.community","slug":"mlops-community","podcast_index_feed_id":28679,"rss_url":"https://anchor.fm/s/174cb1b8/podcast/rss","website_url":"https://mlops.community","image_url":"https://d3t3ozftmdmh3i.cloudfront.net/production/podcast_uploaded_nologo/3809022/3809022-1612190855115-e91f8b881173f.jpg","author":"Demetrios","episode_count":516,"summary":"Relaxed Conversations around getting AI into production, whatever shape that may come in (agentic, traditional ML, LLMs, Vibes, etc)","last_synced_at":null,"page_url":"https://stenobird.com/podcast/mlops-community"},"episode":{"title":"The Future of Information Retrieval: From Dense Vectors to Cognitive Search","slug":"the-future-of-information-retrieval-from-dense-vectors-to-cognitive-search","published_at":"2026-02-17T18:00:11+00:00","page_url":"https://stenobird.com/podcast/mlops-community/the-future-of-information-retrieval-from-dense-vectors-to-cognitive-search","show_page_url":"https://stenobird.com/podcast/mlops-community","url":"https://podcasters.spotify.com/pod/show/mlops/episodes/The-Future-of-Information-Retrieval-From-Dense-Vectors-to-Cognitive-Search-e3f7el9","audio_url":"https://anchor.fm/s/174cb1b8/podcast/play/115636329/https%3A%2F%2Fd3ctxlq1ktw2nl.cloudfront.net%2Fstaging%2F2026-1-17%2F418277708-44100-2-a7817f3055f02.mp3","summary":"Information retrieval is shifting from simple keyword matching to 'Cognitive Search,' where systems reason over retrieved data. This evolution moves beyond dense vectors toward agents that can perform multi-turn actions and personalized reasoning.","meta_description":"Explore the evolution of search from keyword matching to Cognitive Search and RAG with LinkedIn Staff Engineer Rahul Raja.","key_points":["Main idea: The 'R' in RAG (Retrieval) is more critical than the 'G' (Generation); a powerful LLM cannot fix a broken retrieval layer","Practical takeaway: Optimize for cost and latency by accepting higher latency in exchange for better accuracy, especially when using LLMs for reasoning","Failure mode: Relying solely on dense vectors without considering the trade-offs in cost, speed, and accuracy can lead to unsustainable production infra","Main idea: Cognitive Search represents a shift toward multi-turn, agentic retrieval that can perform actions rather than just returning links","Practical takeaway: Use LLMs to enrich metadata and tags (e.g., dietary preferences in menus) to enhance traditional search capabilities"],"chapters":[{"start_ms":65000,"title":"The Criticality of the Retrieval Layer","summary":"Why the quality of your retrieval layer determines the success of RAG, regardless of the LLM used."},{"start_ms":340000,"title":"The Vision of Cognitive Search","summary":"Moving from simple semantic matching to search systems that can execute actions and provide personalized user experiences."},{"start_ms":630000,"title":"Tools for the New Paradigm","summary":"A look at the libraries and infrastructure available for implementing modern embedding-based search."},{"start_ms":930000,"title":"Production Trade-offs: Cost, Latency, and Accuracy","summary":"Navigating the engineering constraints of deploying dense retrieval at scale."},{"start_ms":1215000,"title":"Evaluating Search Techniques","summary":"Analyzing the effectiveness and trade-offs of hybrid search versus pure cognitive approaches."},{"start_ms":1500000,"title":"Optimizing Retrieval Costs","summary":"Strategies for using BM25 and text-based retrieval to minimize expensive vector computations."},{"start_ms":2630000,"title":"The Impact of Embedding Model Changes","summary":"How swapping embedding models can fundamentally alter system performance and downstream results."}],"topics":["Information Retrieval","Retrieval-Augmented Generation","Vector Databases","Semantic Search","Cognitive Search","Machine Learning Operations","Large Language Models","Search Infrastructure"],"duration_seconds":3773,"processing_state":"processed","actions":[{"name":"request_transcript","method":"POST","url":"https://stenobird.com/v1/public/podcasts/mlops-community/episodes/the-future-of-information-retrieval-from-dense-vectors-to-cognitive-search/transcription-requests","description":"Idempotently request low-priority transcript generation for this episode."},{"name":"read_markdown","method":"GET","url":"https://stenobird.com/podcast/mlops-community/the-future-of-information-retrieval-from-dense-vectors-to-cognitive-search.md","description":"Read the agent-friendly Markdown representation of this episode resource."}]}}