The latest news and research from Amazon's science community. #AmazonScience
Global
Joined December 2019
- Tweets7.6K
- Following1.9K
- Followers92K
- Likes10.9K
Amazon Science retweeted
Live from #PyTorchCon 2026, Robert Hundt (@Amazon) introduces TorchNeuron, the native PyTorch backend for @awscloud Trainium.
Robert shows how TorchNeuron supports eager execution and torch.compile, letting developers train on Trainium using familiar PyTorch code and distributed training APIs.
He also covers how vLLM-Neuron uses torch.compile for serving, and how Neuron Explorer lets developers profile execution from PyTorch down to hardware instructions. @AmazonScience
Amazon Science retweeted
We're open sourcing TorchNeuron. Your existing @PyTorch code runs on Trainium with one config change, and we're contributing upstream to make PyTorch more hardware-agnostic for everyone. aboutamazon.com/news/innovat…
Amazon Science retweeted
Trying a new AI chip used to be a lot of work. You often had to rewrite your code, revalidate it, and sink many weeks in, betting this new chip is the right one for your workload.
That cost time and money, and wore on the team. Your best people got buried in plumbing instead of building the thing they came to build.
TorchNeuron is going to largely take that off the table, enabling your existing PyTorch code to run natively on Trainium with one configuration change and no rewrite.
With Trainium’s very appealing price-performance characteristics, TorchNeuron is going to make it even easier for more companies to use AI more cost-effectively. aboutamazon.com/news/innovat…
Amazon Science retweeted
PyTorch runs natively on @awscloud Tranium with no code changes. AWS Tranium is @amazon's portfolio of AI chips designed for next-generation deep learning training and inference workloads. You can use it with the tools and workflows you are already familiar with. No new frameworks to learn. No code to change. During PyTorch Conference North America keynote, Robert Hundt, VP Distinguished Engineer, Annapurna Labs, Amazon will share the story and the capabilities AWS provided to meet performance and quality goals. @AmazonScience
#PyTorchCon
Amazon Science retweeted
Introducing TorchNeuron, a new open-source tool that lets developers take their PyTorch code and run it on @awscloud Trainium chips, no rewrite required.
85% of AI researchers and developers write their code in PyTorch, and now, with TorchNeuron, the tools developers already use to test, debug, and scale their models all work on Trainium without modification. Teams building everything from voice AI to drug-discovery models are already using it. ⬇️
Amazon Science retweeted
More hardware choice for PyTorch developers: TorchNeuron registers @awscloud Trainium as a native device, built on the same extension points the community created to make PyTorch hardware-agnostic.
Open sourcing at PyTorch Conference on Oct. 20. bit.ly/4i3P9Ud
Amazon Science retweeted
At #COLM2026 this week! 👋
I’ll be at the Amazon Science booth for Meet the Scientist — come say hi and chat about LLM post-training, agentic post-training, multimodal learning, world models, video understanding, or anything AI/ML!
📍 Wed, Oct 7: 12–1 PM
📍 Thu, Oct 8: 1–2 PM
We're at @COLM_conf this week! Stop by the Amazon booth for live talks on super weights in LLMs, multi-agent orchestration, responsible AI, and more, plus meet-the-scientist sessions all week. #COLM2026
Full schedule: amzn.to/4jsMYKM
Amazon Science retweeted
Georgia Tech and @AmazonScience have announced a new research partnership. The Georgia Tech-Amazon Hub will focus on developing technologies with practical applications in areas like robotics, logistics, and aerospace. #WeCanDoThat 🐝📦 | b.gatech.edu/3VIBKb9
Amazon Science retweeted
The most durable skill in the age of AI isn’t any single technology. It’s learning.
I was back at @Stanford last week celebrating the Stanford-Amazon Research Initiative, bringing our researchers together to work on hard problems across AGI, robotics, healthcare and more.
Excited to see what we learn and solve together.
Amazon Science retweeted
We spent tens of billions of tokens building a benchmark for AI security — so you don't have to.
Deception Benchmark is now open on GitHub: 14,822 samples, 16 languages, 70+ CWEs.
Download the dataset, run your tools, hold them accountable.
go.aws/4xUbAjP
#AWS #OpenSource
Amazon Science retweeted
"Machine learning, at its core, is about generalization, not memorization," write @awscloud Applied Scientist Martin Bertran Lopez and @WarrenCntrPenn faculty affiliate @Aaroth for @AmazonScience. amazon.science/blog/why-dont…
Amazon Science retweeted
Trainium has by far the best profiler of any accelerator. As you can see we have nanosecond-accurate traces of our programs, allowing you to write a program that generates images reliably in the profiler.
Amazon Science retweeted
Honored to be named to @TIME's 2026 TIME100 AI list. This one belongs to our customers and the teams at @awscloud. Still early.
AWS CEO @MattSGarman has been named to @TIME’s 2026 TIME100 AI list. Congratulations, Matt. go.aws/4gFhle3
Amazon Science retweeted
You discover a task where your LLM is lacking.. you gather expert data and fine-tune for the new skill, only to find it no longer adheres to formatting instructions in simple tasks. How can we prevent this?
I will present a research demo talk at ICML '26 👇🧵
#ICML2026 #ContinualLearning #SelfDistillation
Amazon Science retweeted
📌☁️ @Pinterest just committed $4B for @awscloud services through 2031, the largest infrastructure commitment in the company's history.
Learn how AWS is powering personalized discovery for Pinterest’s 600M+ users with Trainium, Graviton, & Amazon EKS: go.aws/4uJkTSe
Amazon Science retweeted
If a model is 50% reliable, it's 0% useful.
Michael Giannangeli, Head of Agentic AI at Amazon Nova, breaks down one of the hardest problems in agentic AI: reliability.
Models need to match or beat human performance, or they won't get adopted. To close the gap, teams are building an ecosystem of reinforcement learning "gyms" where AI agents can hone their skills. Think of it like a virtual training ground where agents practice tasks through trial and error, building reliability one rep at a time. 💪
