Sully.ai
87 posts
@sullyai
Autonomous OS for healthcare
🩺 @sullyai has returned over 30 million minutes to physicians — more time with patients, less on paperwork. @baseten powers this with their optimized inference stack built using NVIDIA Blackwell, NVFP4, TensorRT LLM, and NVIDIA Dynamo, to run frontier open models like gpt oss 120b. The result: 10x cost reduction and 65% faster responses for workflows like clinical note generation. 🔗 Read the blog: nvda.ws/468smA3
Every healthcare AI co is building for clinicians. Nobody's building for patients. That's the $50B mistake.
@sullyai tested the major STT providers on real MENA clinical audio. 👀 Code-switching, dialect-heavy consultations, the conditions generic models fail on. 🏥 Patrick Nguyen, Head of Engineering MENA: ours was the only one that hit the performance thresholds needed for clinical documentation at regional scale.
Announcing NVIDIA Nemotron 3 Super! 💚120B-12A Hybrid SSM Latent MoE, designed for Blackwell 💚36 on AAIndex v4 💚up to 2.2X faster than GPT-OSS-120B in FP4 💚Open data, open recipe, open weights Models, Tech report, etc. here: research.nvidia.com/labs/nemotron/… And yes, Ultra is coming!
We made an early bet on a full team of autonomous agents. We're #1 on speed! If you walk into any hospital today, you would see them using an average of 50-100 software tools, with some having over 800 SaaS subscriptions! Getting one integration with one solution to your health system at a time is a nightmare. Not just that, getting all those AI tools to talk to each other is nearly impossible. The biggest objection we get before showing people our demo is: how good each and every agent is, how you're doing all those suites of agents, and how you can claim you're better. But we don't like to talk, we show them what we built, and their jaws drop. Why does that happen? because we're driven by UX (user experience). If something is a better UX, our research and engineering team figures out how to do it. If physicians don't want to wait for something, they shouldn't; if technology isn't there yet, we'll figure it out. Thanks to the @sullyai team and our partners for making this happen! PS. A lot of people ask us about the accuracy with this speed, for us, this is clinical information, so quality is out of the question! Read our paper here about our clinical accuracy: arxiv.org/abs/2505.23075
🩺 @sullyai has returned over 30 million minutes to physicians — more time with patients, less on paperwork. @baseten powers this with their optimized inference stack built using NVIDIA Blackwell, NVFP4, TensorRT LLM, and NVIDIA Dynamo, to run frontier open models like gpt oss 120b. The result: 10x cost reduction and 65% faster responses for workflows like clinical note generation. 🔗 Read the blog: nvda.ws/468smA3
🧵 How @baseten, @DeepInfra, @FireworksAI_HQ, and @togethercompute are cutting AI Costs by up to 10x With Open Source Models on NVIDIA Blackwell Across industries, tokens power every AI interaction: 💉 Medical insights 🎮 Game character dialogue 🧙♂️ Autonomous customer support 🤖 Agentic chat
