Hi I'm Bo and I do BO; PhD student LIAC @schwallergroup MSc in CS @EPFL • MSc thesis @ETH AI is your friend (most of the time)
Joined October 2021
- Tweets194
- Following452
- Followers628
- Likes421
Pinned Tweet
🚨New preprint!
✨My precious✨ GOLLuM: GP-Optimized LLMs for Bayesian Optimization — the first of its kind!
LLMs as deep kernels in GPs jointly optimized via marginal likelihood → implicit metric learning, calibrated uncertainty & superior sampling 🚀
📄 arxiv.org/abs/2504.06265
BOjana Ranković retweeted
Very excited about this one: Today we're launching the b12 Foundry in partnership with the High-Throughput Experimentation Lab at the University of Zurich. AI can now design a molecule in seconds, but making it can still take a chemist several weeks. b12 Foundry closes that gap. b12-labs.com/foundry
BOjana Ranković retweeted
Thrilled to share our new work MatInvent, a general and efficient reinforcement learning workflow that optimizes diffusion models for goal-directed crystal generation. Thanks to @JeffGuo__ , @efssh , @pschwllr , @SchwallerGroup , @NCCR_Catalysis .
arxiv.org/abs/2511.03112
BOjana Ranković retweeted
High-dimensional linear mappings, or linear layers, dominate both the parameter count and inference cost in most deep learning models. We propose a general-purpose drop-in replacement with a substantially better capacity - inference cost ratio. Check it out!🧵
BOjana Ranković retweeted
You don’t like molecular dynamics? We get it. That’s why at this year’s LLM hackathon for Chemistry and Materials Science, we built not one, but ✨two✨ AI agents for molecular dynamics 👇
b-12 (b12-labs.com) is building AI agents that help chemists design experiments and automatically run them on lab robots. They turn months of manual chemistry work into minutes of automated execution.
Congrats on the launch @drecmb & @ZJoncev!
ycombinator.com/launches/O3V…
BOjana Ranković retweeted
Heeey!
We are launching four agents on FutureHouse Platform!
Want to give it a try?
🏔️GOLLuM is in the Swiss Alps 😱 but also at the #ICLR2025 World Models workshop! 👇
📍Peridot 201 & 206 🗓 Apr 28th, 12:00
🔗arxiv.org/abs/2504.06265
Fun fact: Tolkien’s mountains were Swiss-inspired. 🧙♂️✨
Dark as I like my coffee ☕ jk, I don't drink coffee but love the dark mode! Never used it for papers tho, thanks for a new experience @Synthical_ai! 🖤 Check out the colors 😻 and let there be (no) light 🧙♂️
🔗synthical.com/abs/2504.06265…
A mandatory 25th of April post 😌
Quick maths 🧮: In a lab of 25 people from 15+ countries, what's the probability that two share the same birthday —🗓️ 25 April— and they're both at #ICLR2025 🇸🇬?
Drop your estimate below, or just wish @6ojaHa 🇷🇸 and @TheoNeukomm 🇨🇭 a huge happy birthday! 🎉🎂🥳🎁🍰
BOjana Ranković retweeted
Hello from Singapore 🇸🇬! Thrilled to be at #ICLR2025 presenting our work on fragment-based drug discovery 🧩. We go beyond virtual screening with a generative, structure-aware approach.
📃 openreview.net/forum?id=bZW1…
🔗 github.com/rneeser/LatentFra…
A thread 🧵👇
BOjana Ranković retweeted
Thrilled to present our new work on Reinforcement Learning for 3D Crystal Diffusion Generation in #AI4Mat at #ICLR2025 ! We show that RL can optimize diffusion models for goal-directed crystal generation. Thanks to @pschwllr and @SchwallerGroup !
openreview.net/forum?id=Ovxf…
How it feels to bash your own previous work in a new paper💀 Anywaaays, hello from #ICLR2025 in Singapore! 🇸🇬
I’m presenting GOLLuM: GP-optimised LLMs at the World Models Workshop on Monday.
Let's discuss LLMs, finetuning, GPs, BO, life, pain, PhD👀
🔗 arxiv.org/abs/2504.06265
🚨New preprint!
✨My precious✨ GOLLuM: GP-Optimized LLMs for Bayesian Optimization — the first of its kind!
LLMs as deep kernels in GPs jointly optimized via marginal likelihood → implicit metric learning, calibrated uncertainty & superior sampling 🚀
📄 arxiv.org/abs/2504.06265
BOjana Ranković retweeted
🔥 AutoAdsorbate 🔥 has been the workhorse of high throughput computation in heterogeneous catalysis modeling and ML @BASF for the previous few years. 🧵
Now it is available open source!
#MatSci #CompChem
Preprint: doi.org/10.26434/chemrxiv-20…
Code: github.com/basf/autoadsorbat…
Thanks to the lovely people at @anaralabs, you can now chat with GOLLuM — the paper, not the framework (yet ⌛️). Ask it about Bayesian optimization, GP kernels, LLM finetuning under uncertainty…
Everyone asks what is BO.
No one asks how is Bo 🥲
anara.com/papers/gollum-gaus…
Congratulations :)
We made it easy for anyone to chat with GOLLuM:
anara.com/papers/gollum-gaus…
Btw, if you're into LoTR, come meet GOLLuM tomorrow! 👀 One representation to rule them all 🙇♀️
I'll be giving a seminar hosted by Green Dynamics on:
📌 Reframing LLMs as Principled Bayesian Optimizers for Chemical Discovery 🧪
📅 April 10, 9:00 AM CET
🔗events.teams.microsoft.com/e…
🚨New preprint!
✨My precious✨ GOLLuM: GP-Optimized LLMs for Bayesian Optimization — the first of its kind!
LLMs as deep kernels in GPs jointly optimized via marginal likelihood → implicit metric learning, calibrated uncertainty & superior sampling 🚀
📄 arxiv.org/abs/2504.06265
Finally, would I trust an LLM with my expensive experiment? Absolutely not. ❌
Would I trust a GP? Yes. ✅
LLM optimized through a GP? Absolutely YES. 💯💯💯
We offer a principled path to using LLMs for real-world optimization — under uncertainty, under low data, under control.
Huge thanks to my advisor @pschwllr for the support (and patience 🙏) and to @NCCR_Catalysis for funding this journey 💙
Code here: github.com/schwallergroup/go…
Paper here: arxiv.org/abs/2504.06265
Future with GOLLuM: less uncertain 🔮
