vespa.ai

Radu Gheorghe is a software engineer at @vespaengine. He joins @seanfalconer to discuss vector search limits, tensor based retrieval, ranking tradeoffs, and where AI search is headed next. @radu0gheorghe softwareengineeringdaily.com/2026/05/12/ves…

Amazing blog post by @vinted team on scaling personalized search autocomplete to 125M suggestions, 4.7k QPS and 31ms p99 latency 🤯

I really look forward to our first @vespaengine full-day community event in London, September 10! Don't forget to submit a talk, or just hang out with other practitioners. Secure your Early-bird ticket now! This is an event to learn from each other and get inspiration for how to build state-of-the-art applications: content.vespa.ai/vespa-live

Self-paced-friendly way to learn @vespaengine at learn.vespa.ai

Blog post: vinted.engineering/2025/11/06/ves…

Vinted explains how to get stable p99 latency of 3ms when using Vespa as a feature store. 👇

Sample app repo: github.com/aws-samples/sa…

AWS has published a sample app which shows how to build a scalable LLM agent grounded in both general and proprietary information. Repo link 👇

Let’s take Vespa from online → real world: content.vespa.ai/vespa-live

It’s happening. Vespa.ai is going live, in person. For the first time ever, we’re bringing the Vespa community together for a dedicated, in-person conference: Vespa.ai Live in London 🇬🇧 If you’ve been building with Vespa, exploring real-time AI, or pushing the limits of search and retrieval, this is the event you don’t want to miss. This isn’t just another conference. It’s the people behind real-world systems, sharing what actually works at scale. 💥 What makes this special: • The first-ever Vespa in-person event • Deep dives into real-time, production-grade AI systems • Talks from practitioners solving hard problems in search, personalization, and RAG • A chance to meet the community face-to-face 🎤 And yes, we want to hear from you. This is your chance to get on stage, share your work, and be part of shaping the Vespa ecosystem. 📍 London: Lumiere London Underwood Flagship Venue 📅 September 10, 2026 Details / submit your talk 👇

Read the the full blog by @bjorncs and @thomas_thoresen where they walk through exactly what's possible, what to watch out for, and where the boundaries are: blog.vespa.ai/onnx-external-…

Big models just got easier to use in Vespa. Many state-of-the-art embedding models, like Multilingual-E5-large or EmbeddingGemma 300M, exceed the 2 GB protobuf limit and store their weights in separate external data files. Until now, that meant they couldn't be used directly in Vespa's built-in embedders. That changes starting with Vespa 8.544. Vespa embedders now support ONNX models with external data when referenced via a model URL. No manual workarounds. No custom scripts. With the external data files colocated with the model, Vespa will fetch and use them automatically, with no custom handling required. This works across: ✅ hugging-face-embedder ✅ colbert-embedder ✅ splade-embedder And if you're on Vespa Cloud, models from the Vespa Model Hub with external data are supported too, including private models with auth token propagation.

This blog post covers the numbers, evaluations, and how to set this up: blog.vespa.ai/asymmetric-ret…

If your application has more queries than writes, you can save a lot by using a small embedding model running locally on cpu at query time. Voyage AI provides a pair of models that lets you do this with state of the art quality, and with Vespa you get to run them out of the box.

Will Vespa be killed by Claude?🤔deathbyclawd.com/?url=vespa.ai

Filters are everywhere in real-world vector search, and they're quietly killing your recall and latency. Join Radu Gheorghe and search veteran Doug Turnbull (formerly Reddit, Shopify, Wikipedia) for a practical deep dive into: ✅ Why HNSW struggles with filtered queries ✅ Strategies like ACORN-1, brute force kNN, post-filtering, overfetching & adaptive beam search ✅ How to use HNSWTuner + VespaNNParameterOptimizer to find your optimal settings Most teams building vector search don't think about this until it bites them in production. This is your chance to get ahead of it. 🔍 Tuning HNSW Parameters for Filtered Search 📅 April 16, 2026 | 5:00 PM UTC, 2PM EST 👉 Register here: maven.com/p/601c82/tunin…

Seriously now, it should be insightful for anyone doing vector search. Ideally with @vespaengine but not necessarily (there are common techniques there). Sign up at maven.com/p/601c82/tunin…

Everyone is talking about AI agents. But the hard part isn’t the agent. It’s retrieval. Agents need to search, filter, rank, and reason over huge amounts of data in real time. Without that foundation, they hallucinate or miss critical signals. In our latest blog, learn how Metal is building agent-driven intelligence on Vespa Cloud, powering AI agents with real-time retrieval and ranking. 👉 Read the story: blog.vespa.ai/agent-driven-i…

#AI introduces new expectations around speed, relevance, and adaptability. In this discussion with experts from @vespaengine, learn how a practical view of the workings of AI in digital commerce reduces complexity by simplifying how systems work together. em360tech.com/podcasts/how-s….

The blog post contains more details and advice: blog.vespa.ai/embedding-trad…

Access the dashboard comparing all eligible embedding models on HuggingFace: huggingface.co/spaces/vespa-e…

Choosing the right embedding model for search. A hard problem because this impacts memory and cpu/gpu usage, latency, and quality. So, we built an interactive dashboard where you can compare all the leading open models. ⬇️