Zilliz Hands-On | Full-Text Search and Vector Search in One Engine - From Hybrid Search to Multimodal Search

🎯 Event OverviewIn this workshop, participants will learn about vector search, hybrid search, and multimodal search using Milvus / Zilliz Cloud, with a focus on Japanese data through hands-on experience. By combining Japanese morphological analysis, BM25, dense vectors, and image search, attendees will gain practical experience in building a search infrastructure that considers performance, operational efficiency, and costs. The second half will include a demonstration of transitioning from an existing Milvus environment to Zilliz Cloud and the new features of Milvus 3.0.Beginners in the field of vector databases are also welcome. We will start together with a free trial registration for Zilliz Cloud on the day of the event. No prior preparations are required; simply bring a laptop and a browser.📍 Venue: TKP Tokyo Station Conference Center, Conference Room 1A 〒103-0028 Tokyo, Chuo-Ward, Yaesu 1-8-16, Shin Makicho Building, 1 minute walk from JR Tokyo Station Yaesu Chuo Exit📆 Date: September 17, 2026 (Thursday) 18:30–21:20 (Registration starts at 18:00)👥 Format: In-person / Hands-on📝 Capacity: 40 people💻 What to bring: Laptop (any operating system as long as the browser works)🍴 Snacks and drinks: Pizza, light snacks, and drinks will be provided.📅 Agenda18:00-18:30 | Registration18:30-19:00 | Session: Operating a Vector Search Infrastructure in Production - Real Solutions for Performance, Cost, and Migration19:00-19:10 | Demo A: Beyond Practical - Complete Japanese Hybrid Search19:10-20:30 | Practical Exercise20:30-20:40 | Q&A20:40-20:50 | Demo B: Transition from Self-Hosted Milvus to Zilliz Cloud - A Hands-On Experience20:40-20:50 | Demo C: Searching While Storing in a Data Lake - Milvus 3.0 External Collections20:50-21:20 | Open NetworkingInstructorYan Yulan / Zilliz Founding Solution Architect, PhDRecommended forHands-on Content (approx. 80 minutes)On Zilliz Cloud, participants will build a Japanese document search from scratch. This will be conducted based on Milvus 2.6.Lab Content TimingLab 0 Environment Setup - Connection Test to Zilliz Cloud Approx. 15 minutesLab 1 Japanese Semantic Search - From Data Collection to Data Retrieval and Search Approx. 15 minutesLab 2 Japanese Word Segmentation and Full-Text Search as well as Hybrid Search Approx. 20 minutesLab 3 Multimodal Search - Retrieving Document Pages with Diagrams Using Japanese Queries Approx. 20 minutesLab 4 (Optional) Advanced Search that is Usable with a Single Added Line Approx. 10 minutesNote: Lab 4 is optional. Even if you do not complete it on the day, you can take the notebook home and try it later in your own environment.Session (30 minutes)Operating a Vector Search Infrastructure in Production - Real Solutions for Performance, Cost, and MigrationThis session will provide insights for those considering vector search and for those already operating it.Vector search has become simpler up to the point of Proof of Concept (PoC). The challenge lies beyond that. As data increases, with no stops to updates, tuning indexes and bills simultaneously becomes a problem – many projects stagnate at this point.In this session, Zilliz, the developer of Milvus, will discuss how it has addressed these challenges. We will cover performance with large datasets using the Cardinal search engine, automation of tuning with AUTOINDEX, cost structures through hierarchical storage and quantization, as well as operations and security, including BYOC. We will also present specific national and international implementation examples and transitions from self-hosted environments.Finally, we will briefly touch on the direction in which Milvus 3.0, released last month, is headed.DemoDemo A: Beyond Practical - Complete Japanese Hybrid SearchDemo B: Transition from Self-Hosted Milvus to Zilliz Cloud - A Hands-On ExperienceDemo C: Searching While Storing in a Data Lake - Milvus 3.0 External CollectionsAbout ZillizZilliz is a company that provides infrastructure for unstructured data for AI and develops the open-source vector database Milvus. As a pioneer in the field of vector databases and vector lakebase, Zilliz helps transform large-scale unstructured data such as text, images, audio, video, code, logs, and user behavior into intelligent assets that can be searched, analyzed, managed, and utilized by AI.Zilliz Official Website⚠️ Notes📩 Inquiries 📧 [email protected]