Friso de Kruiff

PhD Candidate @AmlabUva | Materials Discovery @cusp_ai

[Friso de Kruiff @FrisodeKruiff]

🌊 Flowing with Confidence

Generative models hallucinate citations, produce broken figures, and propose thermodynamically impossible crystals. Validating any of them is computationally expensive.

Can flow-matching models learn to know what they don't know? 🧵

[Max Welling @wellingmax]

Today, @CuspAI is launching the ‘AI Materials Foundry’ - a global network of data, labs, compute, and deep scientific expertise dedicated to the design of novel materials.

[Max Welling @wellingmax]

Today we’re also confirming our $450M Series B funding, valuing CuspAI at $2.6 billion. I am incredibly proud of what our team has built and the scientific rigor they bring to work every single day.

[CuspAI @cusp_ai]

Introducing CuspAI’s AI Materials Foundry.

We’re launching a global network to fundamentally accelerate how we discover and design the physical building blocks of our world.

[Friso de Kruiff @FrisodeKruiff]

@HoldijkLars @sirui_lu97 @wellingmax Love the visuals! Super cool 😇

[Friso de Kruiff @FrisodeKruiff]

Great work by @Mohammad_Niaz94 @rishabh16_ and @Dafidofff! If you’re at @icmlconf go check it out 😇!

[Mohammad Niaz @Mohammad_Niaz94]

Drop by our poster @ICML to learn about scalable equivariant Transformers that retain the speed and simplicity of standard Transformers.

Jul 9, '26 • 5:00 – 6:45 PM, HALL A #2614

Here is the core idea behind #PlatonicTransformers and why the results are exciting.🧵 (1/n)

[Pim de Haan @pimdehaan]

Join us accelerating material discovery at CuspAI! We're looking for for an expert on distributed ML to build the foundation for training & deploying ML force fields.

Please apply at
https://t.co/W5S8w1dJ4N

[Friso de Kruiff @FrisodeKruiff]

One FMwC forward pass matches the rank correlation of Hutchinson's estimator at many probes, at a fraction of the cost.

[Friso de Kruiff @FrisodeKruiff]

🌊 Flowing with Confidence

Generative models hallucinate citations, produce broken figures, and propose thermodynamically impossible crystals. Validating any of them is computationally expensive.

Can flow-matching models learn to know what they don't know? 🧵

[Friso de Kruiff @FrisodeKruiff]

Take crystals as an example. How do we know FMwC's score really reads off the magnitude of the divergence?We compare against Hutchinson's trace estimator on the FlowMM crystals model.

[Max Zhdanov @maxxxzdn]

🌍Today we release Mosaic, a probabilistic weather model that shifts the Pareto frontier of ML weather forecasting.

It matches the skill of state-of-the-art models while generating a 24-member, 10-day global forecast in under 12 s on a single H100.

Thread!