Published inTowards AIEasy Late-Chunking With ChonkieLate Chunking has just been released in Chonkie, a lean chunking library that already boasts over 2,000 stars on GitHub. This is a welcome…5d ago15d ago1
Published inAI Advances8 Common Mistakes in Vector Search (and How to Avoid Them)Vector search looks easy on paper — chuck some embeddings into a database, query them, and boom, you get results. But once you leap from…Jan 28Jan 28
Published inKX SystemsWhy Your Vector Database Should Be Free (If You Have Fewer Than 5M Vectors)A couple of months ago, I gave a presentation on scaling vector databases, and was surprised to find that many engineers using vector…Jan 21Jan 21
Cross-Encoders, ColBERT, and LLM-Based Re-Rankers: A Practical GuideChoosing the right reranking method is challenging, but it doesn’t have to be.Jan 10Jan 10
Vector Databases Are The Wrong Abstraction… Right??Comparing using a vector db and and an existing Postgres databaseJan 101Jan 101
ColPali by Hand ✍︎When retrieving information from large collections of visually complex documents-full of tables, diagrams, and intricate…Dec 20, 2024Dec 20, 2024
Published inKX SystemsModel2Vec: Making Large-Scale Embedding Generation ManageableIf you’ve ever tried generating embeddings at scale-think tens or hundreds of millions of chunks-you know it can be excruciatingly slow…Dec 19, 20241Dec 19, 20241
O1 Pro at $200/Month: Why I’m (Reluctantly) Hooked — and Why Junior Devs Should Be WorriedI’ll be honest: when I first heard that O1 Pro was $200 a month, I nearly laughed it off. Another fancy AI model claiming it can code…Dec 19, 20242Dec 19, 20242
Published inKX SystemsChronos-Bolt: When Speed Meets Accuracy in Time Series ForecastingThe original Chronos model made waves in the time series foundation sphere by showing extremely accurate predictions for a wide range of…Dec 11, 20241Dec 11, 20241
Published inKX SystemsLate Chunking vs Contextual Retrieval: The Math Behind RAG’s Context ProblemIn retrieval-augmented generation (RAG), the quality of document embeddings plays a crucial role in retrieval accuracy. While much…Dec 3, 20243Dec 3, 20243