Open-source vector & data lake to debug, curate and version AI data
Hi HN! I’m Farah, co-founder of Dioptra.ai. We just open sourced katiML (https://github.com/dioptra-ai/katiml) this week and wanted to get your take. katiML is a vector+data lake to debug, curate and version AI data. With katiML, teams avoid the “garbage in, garbage out” effect by taking control over the quality of their data. They quickly and effectively curate high quality data for training, fine-tuning, and fixing hallucinations and edge cases. Features include: - Data Curation: interactive embedding visualization and similarity search, mislabeling and hallucination…
In plain words
katiML is an open-source vector and data lake platform that helps teams manage and improve the quality of their AI training data. It provides interactive embedding visualization, similarity search, and tools to identify mislabeling and hallucinations in datasets. The platform enables data curation for training and fine-tuning models, includes explainability features to detect data drift and bias, and offers an efficient ML data lake for storing and querying vectors and metadata. It's designed for teams building AI systems who want to avoid poor outcomes from low-quality input data.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
Hi HN! I’m Farah, co-founder of Dioptra.ai. We just open sourced katiML (https://github.com/dioptra-ai/katiml) this week and wanted to get your take. katiML is a vector+data lake to debug, curate and version AI data. With katiML, teams avoid the “garbage in, garbage out” effect by taking control over the quality of their data. They quickly and effectively curate high quality data for training, fine-tuning, and fixing hallucinations and edge cases. Features include: - Data Curation: interactive embedding visualization and similarity search, mislabeling and hallucination scores as well as SOTA Active Learning miners. - GenAI 4 explainability: uncover drift and bias in your data with multi-modal foundational models that describe in human language what’s in your data. - ML Data Lake: zero copy, highly efficient ML data lake to store and query any ML data (vectors, metadata etc). - Data Versioning: to track changes in datasets, prompts, and model performance; and pinpoint the causes behind those changes. Check out our github repo(https://github.com/dioptra-ai/katiml), documentation(https://dioptra.gitbook.io/dioptra-doc/EIKhoPaxsbOt062jkPon/) and new slack community channel. Here is a video to show you how we use GenAi 4 explainability(https://www.loom.com/share/c6916272b9914894bddd1a21df0321b4). We'd love to hear about your experiences with iterating on your data to improve models, the challenges you faced, and how katiML could prove useful to your efforts. Let us know what you think we can do better or differently. Looking forward to your comments!
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, June 2023
the whole month →- NI
Heya HN, I've been working on a reddit-like platform as my primary side project for the last few years. Doing a (very) soft launch today, mainly because I want to use it to encourage discussion of alternatives. How non.io works: 1. Free to browse, paid to interact. 2. Minimum subscription is $2 (though you can choose more). I take $1 to run the servers, everything left gets split evenly between everything you upvote that month. It's a simple model, but I hope it's a better one than the freemium model we've been relying on for the last few years. Fundamentally I feel like any ad-supported…
Commerce · 2023 · non.io
Typefully 2.0▲931Effortlessly publish on Twitter and LinkedIn, now with AI
Growth · 2023 · typefully.com
SlashPage▲760Create websites in seconds to build your first 1K community
Growth · 2023 · slashpage.com
