AI PhD at @berkeley_ai | prev at @UniofOxford | ex @Meta
Berkeley, CA
Joined July 2022
- Tweets13
- Following45
- Followers261
- Likes44
Pinned Tweet
AI can optimize materials 🤘
Our (@pabbeel, @svlevine, @AIatMeta) proposed transformer model 𝗖𝗹𝗶𝗾𝘂𝗲𝗙𝗹𝗼𝘄𝗺𝗲𝗿, combined with 𝗲𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻 strategies, discovers materials that optimize target properties.
arxiv.org/abs/2603.06082 – at Berkeley, CA
Optimize materials yourself! 🔥
No cloning, no install — just open our interactive Google Colab demo and start optimizing crystals for formation energy or band gap in < 2 minutes.
→ Try it here: colab.research.google.com/dr…
→ Or grab the repo + open weights: github.com/znowu/CliqueFlowm… huggingface.co/iamkuba/Cliqu…
3/N All weights are fully open on Hugging Face (including the stronger Alexandria-scale checkpoint we trained with more data than in the paper).
Ready for you to use today — whether you’re doing materials discovery, running your own oracles, or just exploring.
Let us know how you like it and which properties do you want to optimize materials for!
🚀 arXiv: arxiv.org/abs/2603.06082 #AIforScience #MaterialsDiscovery #CompChem
2/N The demo is truly plug-and-play.
Run the cells → your crystal becomes a controlable latent vector.
You can even upload your own structures and navigate them inside the model's imagination - the latent space!
3/N How it works:
CliqueFlowmer encodes materials into continuous, fixed-dimensional representations and optimizes them for target properties (e.g., band gap).
It then uses transformers and flow models to decode these optimized latents into new materials.
→ No sampling
→ Direct optimization
2/N Code release
We open-source CliqueFlowmer — a full pipeline for offline materials optimization.
Includes:
• training + inference
• latent optimization (evolution strategies)
• decoding to crystal structures
github.com/znowu/CliqueFlowm…
LLMs that learn without data 🤯
No new datasets. No human demos. Just a model playing against itself—getting smarter every round.
We call it Language Self-Play.
Full paper arxiv.org/pdf/2509.07414
Thanks @metaai and @berkeley_ai
"Better than what's known" - the dream of AI in model-based optimization. Excited to present my paper with @pabbeel & @svlevine on Functional Graphical Models (arxiv.org/pdf/2401.05442) giving theoretical foundations for chasing that dream, at @aistats_conf ! #aistats_conf #ai – at Valencia, Spain
Kuba retweeted
Deep RL has been driven by improvements in handcrafted algorithms. Our NeurIPS 2022 paper, “Discovered Policy Optimisation” instead meta-learns in a space of theoretically-sound algorithms and beats PPO on unseen tasks! w/ @kuba_AI @_aletcher @Luke_Metz @casdewitt @j_foerst 🧵
It takes @PiotrTempczyk , my first AI mentor, and his team to make true breakthroughs in ML; presented gloriously in the paper arxiv.org/pdf/2206.14882.pdf at #ICML2022 in the long-talk fashion!