๐๏ธ The latest episode of Let's Talk Research is now live!
We speak with @sacosta_gut and Leon Wehrhan about MLIPAudit, our open benchmarking suite for MLIP models.
Learn what rigorous benchmarking reveals about reliability, and what trustworthy molecular simulation really means.
"Research would come to us with something brilliant. And then we'd come back to reality and realise we couldn't actually use it."
@ThomasLecat, Head of Growth at InstaDeep
Moving from a lab breakthrough to a real-world product is one of the defining challenges of industrial AI. ๐งต
In Decisive Agents 03: The PCB Issue, we go deep on how InstaDeep is applying reinforcement learning to some of industry's hardest optimisation problems โ from routing trains across live networks to designing printed circuit boards with AI. โก
Brilliant conversations, brilliant energy. Thanks to everyone who came to meet us there, it's exactly these exchanges that remind us why moving decision-making AI from research to real-world impact matters. ๐
Missed meeting our team? Get in touch here: https://t.co/bDBIH65oZm
That's a wrap on Deep Tech Momentum 2026! ๐ฉ๐ช
Berlin delivered, we met brilliant industry leaders shaping the AI of tomorrow. Thank you to everyone who stopped by our Booth! #DTM26
Don't miss Episode 9 of Let's Talk Research!
Catch up now with @sacosta_gut and Leon Wehrhan as they break down MLIPAudit, InstaDeep's benchmark suite for evaluating Machine Learning Interatomic Potentials, so you can be confident on which model works best for your application.
Our team had a fantastic three days learning about recent breakthroughs across cellular therapies, novel targets, emerging tools, and more. Thank you CIMT for having us! ๐
If you're interested in reading the full posters, comment 'poster' below and we'll send you the PDFs!
๐Explore our related work: https://t.co/AgzZ2iWS4H
InstaDeep was honoured to present three posters at #CIMT2026 in Mainz, Germany, Europe's largest event focused on cancer immunotherapy research and development ๐งต
- Expands the explorable immunopeptidome
- Enables database-free exploration of candidate cancer antigens
- Distinguishes spectral signatures between mutated and healthy cells
๐We're excited to release mlip v2, a major upgrade to our open-source machine learning interatomic potential library.
Access faster kernels, a new model architecture, richer physics, and more simulation modes, all in one unified, modular framework ๐งต
With new pretrained models and a detailed migration guide covering all API changes, we're excited to see what the community builds with it.
๐ Docs & migration guide: https://t.co/0v9kbvxAgh
๐ป Code (GitHub): https://t.co/k56VWvALub
๐Paper: https://t.co/ojHiyRbXlG