​Meetup Description:

​Search is evolving fast. From new algorithms and tools to AI-driven solutions, the field is constantly shifting. BASED Meetup is where professionals & enthusiasts discuss the latest trends, breakthroughs and challenges in modern search.

​We dive into topics like:
- Agentic Retrieval-Augmented Generation (RAG);
- Innovative approaches to sparse, dense, and hybrid retrieval;
- Advances in indexing algorithms and vector quantization;
- Modern metrics for search quality;
- Optimization techniques for production-scale systems;
- Cross-lingual and multimodal search challenges;
- The intersection of search and generative AI;
… and much more.

​No sales pitches, no gatekeeping — just an open space to share ideas, learn from each other, and explore the technologies shaping the future of search.

​Whether you're building production systems, researching search algorithms, or just curious about the field, you're welcome here.

​Meetup Info:

​We enjoy experimenting with the meetup's format, and are inspired by you & your feedback! This time, we'll try secret lightning talks, next time it can be an expert panel — let us know what you would like to see!

​Agenda

​18:00 — Doors Open;
18:15 - 18:30 — Welcome Keynote (Evgeniya Sukhodolskaya & Daniel Wrigley, Adrian Hupka - tacto.ai);
18:30 - 18:40 — Lightning Talk 1 (Evgeniya Sukhodolskaya)
18:40 - 19:15 — "The Hitchhiker’s Guide to Vector Databases" (Clelia Astra Bertelli);
19:15 - 19:30 — Break;
19:30 - 19:40 — Lightning Talk 2 (Daniel Wrigley)
19:40 - 20:15 — "Can Images Improve RAG? Insights from Multimodal Experiments in Industry" (Monica Riedler);
20:15 - 21:00 — Networking with Drinks & Pizza:)

​Talk Descriptions:

​Talk 1: The Hitchhiker’s Guide to Vector Databases
Speaker: Clelia Astra Bertelli (LlamaIndex)
Abstract: Vector databases are inherently a challenge, especially if you approach them with the mindset shaped by traditional search. This challenge is not only limited to searching within a database, but extends also to data preparation, ingestion, querying and evaluation: choosing the right embedding model, the most suitable chunking technique, and the best evaluation metrics can give developers serious headaches! 
This won’t be a theoretical speech: it will be a hands-on, practical guide for all the vector databases hitchhikers out there who are interested in optimising their pipelines. You’ll see how I combined hybrid search and reranking to give a Discord bot access to Pokémon knowledge, or how I built an AI agent to learn Go faster, or optimised my queries to make my AI startup advisor deliver precise information, and many other things!  
You’ll get to know the bright side, as well as the deepest, darkest secrets, the pitfalls, the mistakes made - so that, once you get to the outer vector space, you won’t get lost in there :)

Talk 2: Can Images Improve RAG? Insights from Multimodal Experiments in Industry
Speaker: Monica Riedler (PhD Candidate, TU München)
Abstract: Large Language Models (LLMs) like GPT-4 have demonstrated strong performance in question answering, but they often lack domain-specific knowledge and are prone to hallucinations. Retrieval-Augmented Generation (RAG) addresses these limitations by grounding responses in external data, and recent advances in multimodal models, capable of processing both text and images, open up new possibilities.
This talk presents a series of experiments exploring how multimodal models can enhance RAG systems in the industrial domain. We investigated whether incorporating images alongside text improves performance, and compared two strategies: using multimodal embeddings versus generating textual summaries from images. The study evaluated two LLMs for synthesizing answers — GPT-4 Vision and LLaVA — with an LLM-as-a-Judge approach for evaluation.
Our findings show that multimodal RAG can outperform text-only setups, though image retrieval remains more challenging than text. Notably, leveraging textual summaries from images appears more promising than relying on multimodal embeddings, offering greater potential for future improvements.