Dan Zhang @ ICLR (@DZhang50) / X

Dan Zhang @ ICLR

982 posts

Dan Zhang @ ICLR

@DZhang50

LLM Lead at Ricursive Intelligence | ex-Gemini @ Google DeepMind | Computer Architecture PhD @ UT Austin🤘 | Opinions stated here are my own.

SF Bay Area, CA

Joined November 2014

Pinned
Dan Zhang @ ICLR
@DZhang50
Feb 9, 2022
Currently, datacenter ML training and inference uses commodity TPU and GPU devices optimized for a wide range of workloads. Given the extreme scale of large datacenter deployments, would it be practical to build custom accelerators optimized for specific workloads? (1/4)
Dan Zhang @ ICLR
@DZhang50
Jul 21, 2025
lol
291K
Dan Zhang @ ICLR
@DZhang50
Jun 6, 2025
getting tired of winning 😎
Ian Nuttall
@iannuttall
Jun 5, 2025
new gemini 2.5 is out, and cost is so good vs o3 and opus anybody else getting model fatigue? 😂
63K
Dan Zhang @ ICLR
@DZhang50
Jun 11, 2025
🤔
27K
Dan Zhang @ ICLR
@DZhang50
Jun 17, 2025
I'm on my first Gemini paper! 🥰
31K
Dan Zhang @ ICLR
@DZhang50
Aug 25, 2025
We're hiring!
heiner
@HeinrichKuttler
Aug 25, 2025
Sounds like there's a lot of alpha in just hiring the best. I wonder if anyone knows a place that does that?
Careers at Google DeepMind
From deepmind.google
61K
Dan Zhang @ ICLR
@DZhang50
Aug 7, 2025
concerning
23K
Dan Zhang @ ICLR
@DZhang50
Apr 9, 2025
The first tpu that I contributed to has been announced! Excited to see what cool models will be trained and served on this new platform!
Logan Kilpatrick
@OfficialLoganK
Apr 9, 2025
Introducing Ironwood, the first TPU built for the age of inference, and the timing could not be better : ) - Ironwood perf/watt is 2x relative to Trillium, 6th gen TPU - Ironwood offers 192 GB per chip, 6x that of Trillium - 4.5x faster data access blog.google/products/googl…
22K
Dan Zhang @ ICLR
@DZhang50
May 8, 2022
Replying to @ZoeSchiffer
He left a bit after 4 years, which means the real reason is that his stock cliff hit.
Dan Zhang @ ICLR
@DZhang50
Jun 24, 2025
xai has free coffee on weekends; we have free tongsui on weekdays
38K
Dan Zhang @ ICLR
@DZhang50
Jun 12, 2025
can't stop making memes
8.3K
Dan Zhang @ ICLR
@DZhang50
Dec 30, 2023
Replying to @paulg
That's not what the article actually says though. If you read it, it says DEI related job postings dropped by 44% in 2023.
44K
Dan Zhang @ ICLR
@DZhang50
Dec 15, 2022
EA is full of 10 page studies that come to obvious conclusions that they could have figured out by having a 10min conversation with an actual expert in the field 😉
Marius Hobbhahn
@MariusHobbhahn
Dec 14, 2022
We modeled the performance of FET-based GPUs assuming that transistor miniaturization will hit a limit before reaching the size of a silicon atom. Our model predicts that performance will plateau between 2027 and 2035 at ~1e14 to 1e15 FLOP/s in FP32. 1/n epochai.org/blog/predictin…
Dan Zhang @ ICLR
@DZhang50
Jun 3, 2021
I wrote a paper with a few Google colleagues about FAST, a new technique to build new specialized ML hardware accelerators able to improve computer vision inference performance by up to 6x relative to TPU-v3! arxiv.org/abs/2105.12842 (1/5)
arxiv.org
A Full-Stack Search Technique for Domain Optimized Deep Learning...
The rapidly-changing deep learning landscape presents a unique opportunity for building inference accelerators optimized for specific datacenter-scale workloads. We propose Full-stack Accelerator...