About us
We’re a group of search & AI enthusiasts committed to sharing knowledge, tips & tricks, cool tools and techniques. We’ll cover a wide range of search technologies, from OpenSearch to Solr, Weaviate to Vespa - and our talks will be informative and useful, not just product pitches. The rise of AI has changed the world of search - so we want to know how it might solve some of our oldest problems, from data quality to query understanding to conversational interfaces. We’d love to hear about how you used search & AI to solve real business problems, create entirely new ways to access information and supercharge your business. Topics covered at our events will include how to measure & evaluate search & AI applications, indexing techniques, search engine operations, relevance tuning, site search, enterprise search, e-commerce search, scaling, RAG & other generative AI but also old-school search … anything cool, anything interesting, anything Search!
Check out the videos of our past talks .
Hosted by:
- Charlie Hull, The Search Juggler (past host of the London Lucene/Solr Meetup, London Enterprise Search Meetup, Haystack LIVE! Meetup with 25 years experience of the search world).
- Eliatra (providing support, managed hosting and services for OpenSearch)
Supported by:
Upcoming events
1
Vector-Native Faceting for E-commerce & Practical Search Migrations
Lumiere London, 6-14 Underwood St, London, GBWe're back after the summer with a collaboration between the London Information Retrieval & AI Meetup and Vespa AI Live !
Starting at 6.30 pm we'll have three great talks, pizza and drinks and chances to network, chat and ask questions on any aspect of search & AI.
********
As this is a joint event, please fill out this form in order to complete your registration - don't just register on this Meetup group.
********Our speakers are:
1. Philippe Bouzaglou: Introducing Vector-Native Faceting
Philippe pioneered the concept of the Social Graph while attending Harvard with Mark Zuckerberg. He is now the technical founder at Vectra, an AI company creating the foundation model for e-commerce search.
"Search queries have long been a one-shot game. The user types in a query, presses "Search" and hopes to find what they were looking for. This pattern was dictated by technical constraints in the era of lexical search, but with vector search, we can remove this constraint and offer the user a much better search experience, where they can iterate and refine their query as they see the results that each step brings. The technical breakthrough that allows for query refinement in vector search is the creation of "control vectors", which nudge the original query vector after each refinement step to create a final, composite vector incorporating the information from all the query refinements steps. We will demonstrate this technique in action on a e-commerce catalog, where the user can add facets to refine their search."
2. Ravindra Harige : Patterns from Shipped Migrations
Ravindra is the founder of Searchplex, a firm focused on designing and building scalable AI-native retrieval and discovery systems across multiple industry verticals.
"Migrating to Vespa means dealing with systems that have accumulated years of schema decisions, analyzers, filters, query DSL usage, ranking rules, business logic, and product-specific edge cases. The hard part is that these concerns are often tightly intertwined: schema design affects query behavior, query builders encode business rules, and ranking logic depends on both.
This talk shares patterns from shipped Vespa migrations across e-commerce and regulated-domain search systems, with examples involving saved searches and alerts, boolean logic, lexical matching, and vector signals. The talk includes a demo of tooling Searchplex is building to get teams to a working Vespa v0.1 faster."
3. Radu Gheorghe : Making Filtered HNSW Fast Again
Radu is a Software Engineer at Vespa ai
"We'll start by looking at how mutable HNSW works and why it's slow(er) for filtered search. Then we'll dig into ACORN-1 and how it explores the graph more to (hopefully) compute fewer distances. For really restrictive filters, we'll look at three-hop exploration (ACORN-1 is two-hop) and exact kNN. We'll visualize results for real queries."
10 attendees
Past events
5
