Qdrant

📅 Join us on November 12 at 6 PM CET! Agentic RAG is transforming how AI reasons, but it also demands modern, rigorous evaluation. In this session, Dat Daryl Ngo from Arize AI will share a practical blueprint to make agentic systems observable, accountable, and self-improving, with hands-on examples, open-source tools, and real-world insights. Don’t miss this deep dive into the future of AI evaluation! Register here: https://lnkd.in/g6sfwccM

We're excited to highlight Evgeniya Sukhodolskaya, our second guest instructor in the 𝐀𝐈-𝐏𝐨𝐰𝐞𝐫𝐞𝐝 𝐒𝐞𝐚𝐫𝐜𝐡: 𝐌𝐨𝐝𝐞𝐫𝐧 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 𝐟𝐨𝐫 𝐇𝐮𝐦𝐚𝐧𝐬 & 𝐀𝐠𝐞𝐧𝐭𝐬 course, starting November 4th (in 3 days!). Evgeniya Sukhodolskaya is a Dev Advocate at Qdrant (formerly at Toloka and Yandex). She will be giving a guest lecture on "𝐌𝐢𝐱𝐢𝐧𝐠 𝐒𝐩𝐚𝐫𝐬𝐞 & 𝐃𝐞𝐧𝐬𝐞 𝐑𝐞𝐩𝐫𝐞𝐬𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧𝐬 𝐰𝐢𝐭𝐡 𝐦𝐢𝐧𝐂𝐎𝐈𝐋", which is a sparse neural retriever that combines the semantic benefits of embeddings within a precise lexical / keyword matching search approach. Hybrid search is often implemented naively (just running separate lexical and vector searches an merging them together), but in the course, we'll be showing better ways to think of this problem and how to best integrate multiple query approaches. Evgeniya's guest lecture will be a key, deep dive example of this. We look forward to Evgeniya's session! If you haven't already, you've only got until MONDAY to still enroll in Doug Turnbull and Trey Grainger's 𝐀𝐈-𝐏𝐨𝐰𝐞𝐫𝐞𝐝 𝐒𝐞𝐚𝐫𝐜𝐡: 𝐌𝐨𝐝𝐞𝐫𝐧 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 𝐟𝐨𝐫 𝐇𝐮𝐦𝐚𝐧𝐬 & 𝐀𝐠𝐞𝐧𝐭𝐬 course! 𝐂𝐥𝐚𝐬𝐬 𝐬𝐭𝐚𝐫𝐭𝐬 𝐓𝐮𝐞𝐬𝐝𝐚𝐲 11/4! Course Details in the Comments 👇

Big news: Qdrant has been named one of Europe’s top 100 rising AI startups in Sifted’s first AI 100 ranking, sponsored by Next47 🚀 The report spotlights the startups shaping the future of AI across Europe, and we’re proud to be among them! Read the full report: https://lnkd.in/gZnEB97

I'll be speaking this morning at GenAI Nightmares about how spooky things can get when you use AI code editors to build an e-commerce site that uses Qdrant... If you're feeling frightful, register here: https://lnkd.in/gzcc5tri

𝗟𝗲𝘃𝗲𝗹 𝗨𝗽 𝗬𝗼𝘂𝗿 𝗥𝗔𝗚 & 𝗩𝗲𝗰𝘁𝗼𝗿 𝗦𝗲𝗮𝗿𝗰𝗵 𝗦𝗸𝗶𝗹𝗹𝘀! 🎓 with Qdrant!! 🔥 Qdrant just dropped the definitive answer! We've released a complete, structured, and 100% FREE course that moves you from beginner to production-ready. ✅ Completely Hands-ON 🎉 📘 Vector Search Essentials & Fundamentals ⚡ Indexing Strategies & Performance Tuning 🔍 Hybrid Search for Maximum Precision ⚙️ Scaling Architectures & Optimization Techniques 🛠️ Building a Production-Grade RAG Search Engine Just pure, high-value content with real-world exercises If you want to understand the "how" and "why" behind RAG systems but not just the simple setup or using encapsulated stuff like supabase or other, this is where you start. 🔗 Playlist: https://lnkd.in/g2MkihPn #ai #agents #vector #db #post #new #skill #network

🎉 It’s been a year since I became a Qdrant Distinguished Ambassador, and what an incredible journey it has been! Over this year, I’ve learned a lot, met inspiring people — both virtually and in person — and had the opportunity to create and contribute to things I might not have achieved otherwise. 🎉 The best part of this journey was the opportunity to present at #“VectorSpaceDay” in Berlin, an unforgettable experience that deepened my connection with the global AI and open-source community. 💕 A huge thank you to Qdrant for trusting us and for building such a strong open-source ecosystem that empowers innovation and collaboration. 🌟 🚀 Just a teaser: my 150th blog article is coming soon — stay tuned! 👉 #VectorSpaceDay: Vector streaming platform https://lnkd.in/gn3W_5AJ #Qdrant #VectorSearch #OpenSource #AI #Community #RAG #LearningJourney #Berlin

Enabling Long-Term AI Memory with Qdrant Dex Bridge built by Tushin Kulshreshtha introduces a new approach to persistent AI memory by using Qdrant as the vector database for storing and retrieving contextual information across conversations. Built with Python, mitmproxy, and the Model Context Protocol, Dex Bridge captures AI interactions in real time, generates embeddings for each message, and stores them in Qdrant for semantic search. This allows assistants like ChatGPT and Claude to recall prior discussions, insights, and recommendations, creating a seamless, memory-enabled experience across sessions and tools. The project demonstrates how Qdrant can serve as a local, privacy-first memory backbone for developers and researchers working on multi-model AI interfaces. Explore the code: https://lnkd.in/gBY6XbcS Explore the demo video: https://lnkd.in/g84D22Pa

𝗪𝗵𝘆 𝗩𝗲𝗰𝘁𝗼𝗿 𝗦𝗲𝗮𝗿𝗰𝗵 𝗜𝘀 𝘁𝗵𝗲 𝗠𝗶𝘀𝘀𝗶𝗻𝗴 𝗟𝗶𝗻𝗸 𝗳𝗼𝗿 𝗣𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝗼𝗻-𝗥𝗲𝗮𝗱𝘆 𝗔𝗜 𝗔𝗴𝗲𝗻𝘁𝘀 In this article, Qdrant DevRel Thierry Damiba explores how AI agents have evolved from simple bots into systems that can plan, retrieve, act and verify in real-world workflows, and how traditional search tools are no longer sufficient at scale. 𝗞𝗲𝘆 𝘁𝗮𝗸𝗲-𝗮𝘄𝗮𝘆𝘀: 👉 Agents face three major production challenges: memory loss (short- or long-term context), fragmented knowledge (multiple data modalities, silos), and brittleness/scaling issues (latency and precision drop off when moving from prototype to production). 👉 Vector search solves core issues by enabling:  • Real-time memory layer for agents (so they can recall previous steps rather than restarting each time).   • Multimodal support and hybrid search (text + image + audio, combining dense and sparse indexing, semantic + metadata filters).   • Millisecond retrieval performance at large scale (important when an agent executes many sequential searches). 👉 Scaling properly matters: As datasets and workflows grow, you need horizontal sharding, replication, vector quantization, on-disk storage, multi-tenant security (RBAC, API keys), and production-grade deployment options (cloud, hybrid, self-host). 👉 Evaluation and observability are essential: Track metrics like recall@k, latency/cost per task, and implement guardrails/fallbacks (including human-in-loop when the confidence is low). 👉 Integration matters: A vector search engine shouldn’t be a bolt-on afterthought. It must plug into agent-builders, memory systems and workflow orchestration tools so you spend more time refining what the agent does, not building the plumbing. 𝗥𝗲𝗮𝗱 𝗺𝗼𝗿𝗲 𝗶𝗻 𝘁𝗵𝗲 𝗳𝘂𝗹𝗹 𝗴𝘂𝗶𝗱𝗲 𝗵𝗲𝗿𝗲: https://lnkd.in/g4GPC9hR

My short trip to Silicon Valley comes to an end today. It's been an intensive time with over 12 hours spent in the DeepLearning.AI studio, but I have a feeling we'll provide something really valuable. It’s been a real pleasure getting to know and learn from Gian Thomas Wrobel. Thanks for all the support and for every great conversation along the way!