vmoens
@VincentMoens
Followers
1K
Following
2K
Media
82
Statuses
553
Member of tech staff @PeriodicLabs TorchRL maintainer (@torchrl1) at PyTorch. Neuroscience PhD, ex-M.D.
London, England
Joined July 2015
Funny that you should mention open source knowing that people such as @vwxyzjn @DBahdanau or myself are well-known opensource contributors, and a big chunk of the team has come to periodic from such big labs đź« Join us at @pediodiclabs we're hiring!
@GoogleDeepMind has many in-house projects that dedicated multi-billion dollar startups are pursuing as their sole business line. Here, GDM is doing what @periodiclabs is doing, but with some differences. Considering that many past breakthroughs from GDM—such as AlphaZero,
1
0
14
the handful of meta folks I’ve spoken to impacted by the layoffs in the last 24 hrs are exceptionally motivated and talented I’m going to make more researcher office hours time this week for anyone affected you can sign up below if that would be helpful to you
21
18
178
This cannot be happening I must be in the wrong timeline
Several of my team members + myself are impacted by this layoff today. Welcome to connect :)
0
0
5
Truly impressed by how well designed this is. It’s cohesive, well structured and simple also they nailed down weight update like no one else!
Today Meta announced torchforge, a brand-new PyTorch-native library that makes it easy to use reinforcement learning (RL) to train AI agents. Forge provides high-performance building blocks and ready-to-use examples, so you can focus on what’s novel about your use case rather
2
0
49
as of 5 minutes ago, our gpt-oss implementation is merged into torchtitan! thanks to all the work by @jianiw_wang @__tianyu at @pytorch making it clean & scalable for the community ❤️ i hope y'all play around with training gpt-oss, it's great for its sparsity & reasoning
periodic ❤️ open-source! for example, we’ve been collaborating with the @PyTorch team to build the highest MFU gpt-oss training implementation (includes thinky sinky flexattn) here’s a few SFT runs of gpt-oss-20b & 120b, where i get ~24% MFU for 20b and ~8% for 120b
9
7
195
I'm in SF for the PyTorch conference + GPU Mode hackathon Reach out if you'd like to meet!
0
0
9
Exciting news! These guys have an A-team, can't wait to see what they come up with
Today we're sharing the next phase of Reflection. We're building frontier open intelligence accessible to all. We've assembled an extraordinary AI team, built a frontier LLM training stack, and raised $2 billion. Why Open Intelligence Matters Technological and scientific
0
0
14
This take is paradoxically very far from the truth Because all the heavy ops are written in other languages the python overhead is very small if not negligible It’s a surprisingly good balance between hackability and perf
its an ironic twist of fate that the most performance intensive workloads on the planet running on eye wateringly expensive hardware are run via one of the slowest programming languages with a precarious parallelism story
0
0
2
I don’t want to be that kind of guy but number of downloads of a package is arguably the worst metric of adoption you can imagine
1
0
4
I think people underestimate this: most of the advances we're seeing wouldn’t exist without open-source + open research. Take away PyTorch or HuggingFace and the whole thing collapses. AI the most awe-inspiring collective effort of our time.
periodic ❤️ open-source! for example, we’ve been collaborating with the @PyTorch team to build the highest MFU gpt-oss training implementation (includes thinky sinky flexattn) here’s a few SFT runs of gpt-oss-20b & 120b, where i get ~24% MFU for 20b and ~8% for 120b
2
6
36
periodic ❤️ open-source! for example, we’ve been collaborating with the @PyTorch team to build the highest MFU gpt-oss training implementation (includes thinky sinky flexattn) here’s a few SFT runs of gpt-oss-20b & 120b, where i get ~24% MFU for 20b and ~8% for 120b
21
21
487
This is the best thing Well done thinky!
Introducing Tinker: a flexible API for fine-tuning language models. Write training loops in Python on your laptop; we'll run them on distributed GPUs. Private beta starts today. We can't wait to see what researchers and developers build with cutting-edge open models!
0
0
8
In med school I realized you can pick up 90% of neurosurgery in 2 weeks. The hard part isn’t the routine, it’s the 1% of times when an artery blows and the room turns into a horror movie. You can be decent at lots of things, but you can’t crush everything
I didn’t start doing physics until I was 33 and now I work on a cutting edge research project at one of the top labs in the world. You can just do things
0
0
3
AI needs to be connected to the physical world, proud to be supporting !
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
7
11
191
Liam was one of the original few who created ChatGPT and now he and team have taken on the challenge of using machine intelligence to accelerate Moore’s law, and scientific research in general at the new @periodiclabs
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
23
46
826
I got Gallium I think it's because people think I'm French 🤨
On your first day @periodiclabs you get to pick an element from the periodic table for your desk plate
0
2
35
excited to share what we’ve been up to! at Periodic Labs, we’re creating AI scientists they think, design experiments, and optimize for results they also run experiments by moving robot arms in an autonomous lab it’s a fun, multi-turn task, with lots of open problems for AI
periodic.com
From bits to atoms.
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
8
32
195
I joined @periodiclabs in May. We’re building AI scientists + autonomous labs to create breakthroughs you can hold. Longer take: https://t.co/g4kMSiwYBP More on Periodic:
periodic.com
From bits to atoms.
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
18
23
347
The team @periodiclabs is absolutely cracked. Favorite team I've ever worked with. We're building AI scientists, and our progress is breakneck - already in the physical lab! The world doesn't need another social media app. We're turning bits into atoms.
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
20
48
850
Super excited to share that I joined Periodic Labs ~4 months ago! At OpenAI, I witnessed first-hand as RL broke benchmark after benchmark, and felt the big AI question increasingly shift from "how do we optimize?" → "what do we optimize?" To me, the answer is clear:
Today, @ekindogus and I are excited to introduce @periodiclabs. Our goal is to create an AI scientist. Science works by conjecturing how the world might be, running experiments, and learning from the results. Intelligence is necessary, but not sufficient. New knowledge is
10
20
348