Brian Moore

𝟏𝟏 𝐫𝐨𝐛𝐨𝐭𝐢𝐜𝐬 𝐟𝐨𝐮𝐧𝐝𝐞𝐫𝐬 𝐰𝐚𝐥𝐤 𝐢𝐧𝐭𝐨 𝐚 𝐫𝐨𝐨𝐦. Here's what they had to say about the future of physical AI. Dive into 𝐁𝐞𝐬𝐬𝐞𝐦𝐞𝐫 𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐬 → bvp.com/atlas/bessemer…

i like comfyui and i love fiftyone so i smashed them together and made FiftyComfy

2025 had 261 work days. in that time i shipped 116 integrations for fiftyone. 66 datasets 38 models 12 plugins that's not to mention the various workshops, virtual events, and in-person meetups held around the world in places like ann arbor, boston, chicago, munich, berlin, dusseldorf, paris (where i got to meet @mervenoyann, @reach_vb, and others from hf) amsterdam (where i got to meet @tuanacelik) , brussells (where i got to meet @NielsRogge), stuttgart, saarlands, and more places that i probably forgot here's a quick summary of what i focused on: --gui agents-- this was the year gui agents took off. i built the data infrastructure to support it—17 GUI grounding datasets, 6 visual agent models (GUI Actor, ShowUI, UI-TARS, OS-Atlas, MiMo-VL), plus tooling to collect, synthesize, and evaluate gui data inside fiftyone if you're training agents that see and click, you need to debug what they see. --document visual ai-- enterprise wants multimodal RAG that works on real documents. i integrated ColPali, ColQwen, Jina v4, the ModernVBERT variants—alongside ocr models and document datasets spanning forms, receipts, and scanned text. fiftyone now handles visual document retrieval end-to-end. --plugins-- datasets and models are useless without workflows. i built a gui dataset collector for capturing and annotating screen interactions in coco format. a lerobot importer that preserves multi-camera views and trajectory metadata. a wandb plugin for tracking training data and model predictions with full lineage. text evaluation metrics (ANLS, CER, WER) for benchmarking OCR. NVIDIA NeMo Retriever Parse for extracting structured text with bounding boxes. the plugins close the loop between data curation, model evaluation, and experiment tracking. --vision language models-- the vlm landscape moves fast. I kept fiftyone current: Qwen2.5-VL, Florence2, Kimi-VL, PaliGemma2, MedGemma, Nemotron Nano, FastVLM, MiniCPM-V, Moondream3, Qwen3VL. each one integrated into the remote model zoo so you can run inference on your data in a few lines. whatever vlm you're using, it should work with your data tooling. --physical ai-- early investment in physical AI. built a lerobot dataset importer and started exploring how fiftyone can support policy evaluation and failure analysis for robotic manipulation tasks. The importer handles multi-camera views, episode grouping, and full trajectory metadata—joint states, actions, velocities, efforts. where things are heading in 2026? and how can i make fiftyone useful for where visual AI is actually going? i have a hunch that vision language action models is the next wave

Data annotation has been a blocker for CV projects... until now. This article, thanks to @Voxel51, explores how the Verified Auto Labeling (VAL) workflow combines zero-shot models with human-in-the-loop QA to dramatically cut labeling cost and time. towardsdatascience.com/computer-visio…

Better DATA = Better MODELS! But the biggest bottleneck in computer vision is data annotation. It's expensive and slow. Introducing Verified Auto Labeling. An AI-assisted approach that uses foundation models like YOLOE to match human quality and speeds up object detection.

Our biggest release yet! 🚀 Evaluating models interactively alongside your data is a powerful technique for understanding their weaknesses and identifying the right actions to take to refine your data and models. Now available natively in FiftyOne 1.1 and 100% open source! 🙌

It's official...FiftyOne Open Source 1.0 is here! 🚀 This is a major milestone for AI builders working with visual data. With over 2M installs, FiftyOne has been making life easier for developers and this new release takes it even further. Here’s what’s new: ➡️ Open source ML techniques: FiftyOne Brain is now fully open source, letting you analyze and tweak your datasets and models with ease. ➡️ Modal panels: Explore individual data samples and scenes with new "micro" plugins, adding flexibility to how you work with your data. ➡️ Change tracking: All FiftyOne datasets now have automatically populated creation/modification fields on their samples, giving you a system of record for your visual AI data. ➡️ Remote zoo models: Download and run custom models whose definitions are in GitHub directly in FiftyOne, expanding on the hundreds of natively available models and existing integrations with popular frameworks like Hugging Face and Ultralytics. ➡️ Remote zoo datasets: Use new native support to load datasets whose download and preparation instructions are stored in GitHub. We can’t wait to see what you’ll build! 👉 Learn more here: voxel51.com/press/voxel51-… #OpenSource #MachineLearning #VisualAI #AI #FiftyOne

📢 FiftyOne 0.25 is here! Now with Python Panels, Custom Dashboards, SAM 2 & Elasticsearch Integration, this release is packed with game-changing features that take your visual AI development to the next level. Here's a peek at what's new: ✨ Python Panels: Build custom App panels effortlessly with a powerful Python interface. Access interactive plots, tutorials, and workflows that are tailored to your needs. ✨ Custom Dashboards: Create dashboards to display stats of interest and quickly gain insights into your datasets. ✨ SAM 2: Say goodbye to manual labeling! The latest Segment Anything model (now available in FiftyOne!) can auto-segment objects in image and video datasets, saving you time and effort. ✨ Elasticsearch Integration: Supercharge your vector search by connecting FiftyOne to your Elasticsearch cluster for scalable text and image similarity searches. ➡️ Learn more and give FiftyOne 0.25 a test drive here: voxel51.com/blog/announcin… #ComputerVision #AI #VisualAI #MachineLearning #DataScience #FiftyOne

We’re celebrating four years of open source FiftyOne! Thank you to our incredible community. Open source software like this doesn’t happen without the support of an amazing community behind it. 💪 🙌 voxel51.com/blog/four-year… #VisualAI #AI #ComputerVision #MachineLearning #OSSCommunity #FiftyOne

FiftyOne 0.24.0 is here, now with full native support for 3D scenes, including 3D meshes, 3D geometries & point clouds! Plus, the new custom workspaces feature lets you save your favorite layouts & load them later with a click. And more! Get the details: hubs.ly/Q02y-tRV0

Excited to partner with @LindseyLi_, @davidcowan, and the rest of the incredible team at @BessemerVP!

Thanks for sharing! Understanding multimodal data, including text, language, and #visual data, is absolutely essential to building intelligence that can reason about the physical world💡

Particularly excited to welcome @LindseyLi_ and @davidcowan to the board. Can't wait to continue building together! 💪

We're incredibly grateful for the support of everyone who’s helped us reach this milestone—our investors, advisors, open source #FiftyOne community, customers, and partners. Each and every contribution has helped us get where we are today, and we're thankful for you!

Thrilled to announce our Series B led by @BessemerVP! 🚀🚀🚀 At @Voxel51, we're on a mission to make visual AI a reality, and this milestone marks an exciting inflection point in our journey. Check out my blog post to learn more: voxel51.com/blog/announcin…

Starting a company is wild, adversity-filled journey! The @Voxel51 story is no different. Stay energized. Stay humble. And roll up your sleeves. Check out @anniegarofal0's How I Met My Co-Founder episode with my co-founder, @brian_e_moore, and I talk through the early days of building Voxel51 into the open source powerhouse for #VisualAI it is today. podcast.confidante.info/e/ep-17-%e2%80…

#FiftyOne just crossed 2 Million downloads. What an amazing feat for @Voxel51. And impact for visual AI! Thank you to all of our users and contributors. github.com/voxel51/fiftyo… Or pip install fiftyone Now a question: how should academic founders include downloads on a vita? Is there a mapping from downloads to citations?!?!

Great to see @Voxel51's partnership with @MongoDB in the #GenAI news! #Atlas and #FiftyOne are a perfect match for building visual AI applications, which require organizing and computing on vast amounts of unstructured data, now featuring our Altas vector search integration!

@zem42 Consider joining us at @Voxel51! We maintain an open source package called fiftyone. It’s like pandas for visual AI and 100% Python with lots of interesting data-centric work

me, mourning my job over on hackernews. also shout out to my teammates and manager, all of whom were amazing to work with. looking for work in the sf bay area, particularly anything programming language development or tooling related