Vikram (@msharmavikram) 's Twitter Profile
Vikram

@msharmavikram

@NVIDIA Sr. Research Scientist | UIUC PhD
All opinions and tweets are personal.

Bluesky: msharmavikram.bsky.social

ID: 17336962

linkhttp://msharmavikram.github.io/ calendar_today12-11-2008 10:39:05

520 Tweet

1,1K Followers

566 Following

NVIDIA AI Developer (@nvidiaaidev) 's Twitter Profile Photo

📣 Introducing NVIDIA Dynamo, a high-throughput, low-latency #opensource inference library for deploying AI reasoning models in large-scale distributed environments. Boost the number of requests served by up to 30x with DeepSeek-R1 on NVIDIA Blackwell ➡️ nvda.ws/3DYHc11

📣 Introducing NVIDIA Dynamo, a high-throughput, low-latency #opensource inference library for deploying AI reasoning models in large-scale distributed environments. 

Boost the number of requests served by up to 30x with DeepSeek-R1 on NVIDIA Blackwell ➡️ nvda.ws/3DYHc11
Vikram (@msharmavikram) 's Twitter Profile Photo

Congratulations, Kyle and team from Avian.io 🚀🚀🚀 Kyle and I started discussing inference about a year ago, and I am super glad to see what they have achieved. Way to go, and more to come!

Hao AI Lab (@haoailab) 's Twitter Profile Photo

We are beyond honored and thrilled to welcome the amazing new NVIDIA DGX B200 💚 at Halıcıoğlu Data Science Institute Hao AI Lab. This generous gift from NVIDIA is an incredible recognition and an opportunity for the UCSD MLSys community and Hao AI Lab to push the boundaries of AI + System research. 💪

We are beyond honored and thrilled to welcome the amazing new <a href="/nvidia/">NVIDIA</a> DGX B200 💚 at <a href="/HDSIUCSD/">Halıcıoğlu Data Science Institute</a>  <a href="/haoailab/">Hao AI Lab</a>.
This generous gift from <a href="/nvidia/">NVIDIA</a> is an incredible recognition and an opportunity for the UCSD MLSys community and <a href="/haoailab/">Hao AI Lab</a> to push the boundaries of AI + System research. 💪
Vikram (@msharmavikram) 's Twitter Profile Photo

Super pumped to see this. We have worked closely with the SGLang team on this for several weeks! This is just the start, and there is much more optimization to enable. 🚀🚀🚀

Vikram (@msharmavikram) 's Twitter Profile Photo

I always wanted to do this! Megakernels and warp-level task stealing unlock optimizations beyond CPU-driven execution; despite programming challenges, fully native GPU execution offers a significant edge! Kudos in showing this!