Ion Stoica
@istoica05
Followers
3K
Following
2
Media
2
Statuses
33
Professor at UC Berkeley, co-founder of Databricks, Anyscale, Conviva.
Bay Area
Joined April 2015
RT @Agentica_: 🚀 Introducing DeepSWE 🤖: our fully open-sourced, SOTA software engineering agent trained purely with RL on top of Qwen3-32B.….
0
65
0
Taking a step towards building a modular RL framework with our SkyRL project.
✨Release: We upgraded SkyRL into a highly-modular, performant RL framework for training LLMs. We prioritized modularity—easily prototype new algorithms, environments, and training logic with minimal overhead. 🧵👇.Blog: Code:
4
9
58
RT @haoailab: [Lmgame Bench] o3-pro: A Milestone in LLM Gaming! 🕹️. The leap from o3 to o3-pro is bigger than you might have thought. We te….
0
109
0
RT @uccl_proj: 1/N 📢 Introducing UCCL (Ultra & Unified CCL), an efficient collective communication library for ML training and inference, o….
0
17
0
RT @robertnishihara: The AI compute software stack consists of 3 specialized layers:. 🔧🔧🔧 Layer 1: Training & Inference Framework (PyTorch….
0
20
0
RT @slimshetty_: ✨ NEW SWE-Agents BENCHMARK ✨. Introducing GSO: The Global Software Optimization Benchmark. - 👩🏻💻 100+ challenging softwar….
0
28
0
RT @NovaSkyAI: 1/N Introducing SkyRL-SQL, a simple, data-efficient RL pipeline for Text-to-SQL that trains LLMs to interactively probe, ref….
0
32
0
RT @sumanthrh: Some of our interesting observations from working on multi-turn text2SQL: .- Data-efficient RL works pretty well: We did ver….
0
4
0
RT @andykonwinski: If you had 15min to tell thousands of Berkeley CS/Data/Stats grads what to do with their lives, what would you say? Last….
0
31
0
RT @percyliang: What would truly open-source AI look like? Not just open weights, open code/data, but *open development*, where the entire….
0
194
0
RT @alighodsi: I am super excited to announce that we have agreed to acquire Neon, a developer-centric serverless Postgres company. The Neo….
0
74
0
RT @LakshyAAAgrawal: Real world AI pipelines are often compound, multi-module, and multi-step programs—unlike most RL/GRPO implementations….
0
15
0
🙏.
Rest in Peace to my father @horowitz39 . We love you Dad. David Joel Horowitz was born January 10, 1939 to Phil and Blanche Horowitz in Queens, New York. David was raised in Queens along with his sister Ruth. Phil and Blanche were high school teachers and members of the
0
0
1
RT @lmarena_ai: We thank the authors' for their feedback. However, there are a number of factual errors and misleading statements in this w….
0
32
0
RT @melissapan: 🚨 Why Do Multi-Agent LLM Systems Fail? ⁉️.🔥 Introducing MAST: The first multi-agent failure taxonomy - consists of 14 failu….
0
56
0
RT @skypilot_org: What a night! Huge thanks to everyone who came out to our first SkyPilot meetup — a packed house of builders and insightf….
0
5
0
This journey has been a blast, and I'm very much looking forward to an exciting future, driven by our incredible community.
Around this time 2 years ago, the community helped us launch the very first Arena leaderboard!. Today we’re publishing a blog to celebrate everything we’ve built together on LMArena! 🥳👏. Highlights:.☑️ 3M+ community votes.🤖 400+ models ranked across text, vision,
3
6
100
RT @vllm_project: perf update: we are continuing to see benefits with vLLM V1 engine’s highly performant design. on 8xH200, vLLM leads in t….
0
41
0
One of the underrated strengths of this framework is its suitability for production deployment—offering both reliability and reduced hallucinations. This work builds on concepts from the SPECIFICATIONS paper (, with the Ember Operator abstraction at its.
Ember: an inference-time scaling architecture framework. 🧵 . (1/8)
2
17
83