Ying Sheng (@ying11231) / X

Ying Sheng

1,417 posts

Ying Sheng

@ying11231

Cofounder & CEO @radixark @lmsysorg | @sgl_project (sglang.io) | Do it anyway | Be the light

Joined April 2021

Pinned
Ying Sheng
@ying11231
Dec 8, 2025
We've been running @radixark for a few months, started by many core developers in SGLang @lmsysorg and its extended ecosystem (slime @slime_framework , AReaL @jxwuyi). I left @xai in August — a place where I built deep emotions and countless beautiful memories. It was the best
radixark.com
Ship AI for All
Ship AI for All
549K
Ying Sheng
@ying11231
Jul 30, 2025
Think about the problems in full stack, you’ll be unstoppable. No boundaries, no barriers, do it anyway.
Elon Musk
@elonmusk
Jul 29, 2025
This false nomenclature of “researcher” and “engineer”, which is a thinly-masked way of describing a two-tier engineering system, is being deleted from @xai today. There are only engineers. Researcher is a relic term from academia.
14M
Ying Sheng
@ying11231
Dec 11, 2024
We are hiring for inference at xAI!!! In addition to experienced inference optimization experts, we are also seeking product engineers who’d like to help with connecting inference backend with product, including client api integrations and deployment.
Lianmin Zheng
@lm_zheng
Dec 10, 2024
We released a new image generation model that was trained from scratch at xAI! It has been fun optimizing the inference for it. Because it's an autoregressive model, we can easily port it to SGLang and reuse all the great optimizations like the new zero-overhead batch scheduler,
179K
Ying Sheng
@ying11231
Jul 11, 2025
Shout out to every single person that has been at xAI!
Guodong Zhang
@Guodzh
Jul 11, 2025
Grok4 is such a good model. Shout out to the whole reasoning team and everyone made it happen.
40K
Ying Sheng
@ying11231
Feb 18, 2025
Feel lucky to work with the people at @xai !
Arena.ai
@arena
Feb 18, 2025
BREAKING: @xai early version of Grok-3 (codename "chocolate") is now #1 in Arena! 🏆 Grok-3 is: - First-ever model to break 1400 score! - #1 across all categories, a milestone that keeps getting harder to achieve Huge congratulations to @xai on this milestone! View thread 🧵
31K
Ying Sheng
@ying11231
Nov 1, 2024
Another angle: What I always encourage people is to believe there is no barrier on whatever you want to work on. Whatever training or inference, nothing is really hard if you spent time and focus. They are all paper tigers. The secret is to just ignore people who told you it’s
Stas Bekman
@StasBekman
Oct 31, 2024
Future ML specialization: Inference or Training? Very soon training LLMs will become a domain of a few companies and there will be very little need in experts in LLM training. Especially when LLMs will be at the level of CV cats-vs-dogs quality. Inference expertise on the other
65K
Ying Sheng
@ying11231
Sep 22, 2025
Deterministic inference, here you are. True on-policy RL is on the way. Although we are mostly using off-policy, having a deterministic mode will make many things easier!
LMSYS Org
@lmsysorg
Sep 22, 2025
SGLang now supports deterministic LLM inference! Building on @thinkymachines batch-invariant kernels, we integrated deterministic attention & sampling ops into a high-throughput engine - fully compatible with chunked prefill, CUDA graphs, radix cache, and non-greedy sampling. ✅
46K
Ying Sheng
@ying11231
May 14, 2025
OMG, that’s a very irresponsible suggestion. Do PhD if you want to be a professor, because those people define the game, so the game rule is like that. Or if you want to have some time to purely enjoy research and work for yourself, in case you find a good advisor. Otherwise, do
Hieu Pham
@hyhieu226
May 13, 2025
When I was thinking whether I should do a PhD or just work at Google, the advice I heard the most was: "Do the PhD; you can always come back to Google." Wonder if that is still true today, with Google replaced by any frontier AI lab 🤔
77K
Ying Sheng
@ying11231
Jul 30, 2025
Haha, actually I always tell people researcher equals to engineer.
Hieu Pham
@hyhieu226
Jul 29, 2025
- xAI researcher
47K
Ying Sheng
@ying11231
Dec 13, 2024
How did you know I love coding 😍
2.3M
Ying Sheng
@ying11231
Aug 11, 2025
The open source RL framework Slime(+SGLang) has been validated to train 300+B models with agentic, coding, and reasoning capabilities, efficiently.
LMSYS Org
@lmsysorg
Aug 11, 2025
Honored to see SGLang adopted in RL training for GLM-4.5 at @Zai_org — large-scale validation from a frontier AI lab pushing the boundaries of LLMs!
26K
Ying Sheng
@ying11231
Sep 20, 2025
About scaling, xAI is consistently bet on one direction. One eventually gets successes after repeated tries but conditioned on 100% diligence each time. Successiveness it is.
Dustin Tran
@dustinvtran
Sep 19, 2025
I departed Google DeepMind after 8 years. So many fond memories—from early foundational papers in Google Brain (w/ @NoamShazeer @ashVaswani @lukaszkaiser on Image Transformer, Tensor2Tensor, Mesh TensorFlow) to lead Gemini posttraining evals to catch up & launch in 100 days, then
42K
Ying Sheng
@ying11231
Sep 29, 2025
I have been waiting for a model with properly trained sparse attention for a long time.
LMSYS Org
@lmsysorg
Sep 29, 2025
🎉 Congrats to the DeepSeek team on the amazing release of Sparse Attention (DSA) in V3.2! This fine-grained design sets a new bar for long-context efficiency 🚀 We’re proud that SGLang is an official inference framework for DeepSeek-V3.2 — with optimized sparse attention
43K
Ying Sheng
@ying11231
Apr 6, 2025
I can’t believe how lucky I am to have met the awesome engineers at SGLang. The team jumped in fast and worked hard day and night, with people in different time zones, to get the job done well. More improvements are coming, and I’m touched by how much passion we all share. ❤️
zhyncs
@zhyncs42
Apr 6, 2025
Within 12 hours of Llama 4’s release, we optimized its performance to lead other engines by 20%. I’m proud of SGLang’s exceptional engineering and grateful for its vibrant community. More optimizations are on the way—stay tuned!🚀
17K