Videos
Conference talks and recorded presentations on search & information retrieval — the practitioner knowledge that lives in slides and hallway war stories rather than papers.
-
Max Irwin - The Search Engine Migration Circus — Max Irwin (OpenSource Connections), Haystack Live; a migration playbook (“Hello Search”, feature parity, the “damage” metric) plus real war stories. Companion to Migration between Search Engines.
-
Choosing Indexes for Similarity Search (Faiss in Python) — James Briggs (Pinecone); hands-on FAISS tutorial comparing Flat, LSH, HNSW, and IVF on Sift1M, with benchmarks and index-selection guidance. Companion to Nearest Neighbor Indexes for Similarity Search.
-
Rene Kriegler - Query Relaxation — Rene Kriegler (OpenSource Connections); reframes Query Relaxation as a query recommendation problem and compares word-shape heuristics, term frequency, Word2Vec and a neural network for predicting which query term to drop. Companion to Query Relaxation.
-
Roman Grebennikov - Personalizing Search Results in Real-Time — Roman Grebennikov (Findify); MICES 2019 war stories on real-time LTR personalization: Position Bias feedback loops, the ~1% shuffled exploration segment, one generic cross-merchant model, and purchase-weighted “perfect rankings” (the Stanley bong story). Companion to Exploration vs Exploitation and Position Bias.
-
Evgeniya Sukhodolskaya - Fine-Tuning Sparse Neural Retrievers for E-Commerce — Evgeniya Sukhodolskaya (Qdrant); MICES 2026 on making SPLADE fine-tuning approachable: why off-the-shelf sparse models trained on MS MARCO misfire on catalogs, the ANCE loop that puts a search engine inside training, and the specialize-vs-generalize trade. Companion to Fine-Tuning Sparse Embeddings for E-Commerce Search and Hard Negative Mining.
-
Evgeniya Sukhodolskaya - Relevance Feedback Inside the Search Engine — Evgeniya Sukhodolskaya (Qdrant); Berlin Buzzwords 2026 on the one search component nobody adapts — the scoring function. Model-generated Relevance Feedback steers HNSW hop selection instead of reranking a top-k, distilling a reranker into the index. Companion to Relevance Feedback and HNSW.
Podcasts
- Vector Podcast — Dima Kan (Aiven); interview series running since October 2021, 38 episodes. A primary-source timeline of the vector search wave — the builders of Weaviate, Milvus, Pinecone and Qdrant interviewed mid-build, plus HNSW co-author Yury Malkov and on-site batches from Berlin Buzzwords. Companion to Search Communities.
Related
- Case Studies · Topics · Concepts · Tools · People