Radical AI

History is defined by the materials we master. But the materials of the past cannot build the future. Radical AI is building scientific intelligence to autonomously discover and manufacture next-generation materials to solve the world's greatest challenges.

Three ways to win: 1. Do more than anyone else in your space 2. Know more than anyone else in your space 3. Build a thesis no one else has been able to come up with Our CEO @josephfkrause on the one thing every great entrepreneur has in common and why you need to learn it.

Alloys: We started here because it's hard.

Most ML approaches to materials prediction take the same shortcut: map composition directly to property. It's faster. It's cheaper. It fails in the lab. A material's properties aren't determined by composition alone. They're determined by structure: how atoms arrange in three-dimensional space. Two alloys with identical compositions can behave completely differently depending on element distribution across the crystal lattice. Prediction isn't progress until it's proven in the lab. Our post on why we factor in structure, and what it takes to compute properties that actually hold when synthesized.

Something we deeply believe @RadicalAI 👀 The messiness of real experimental data is a feature, not a flaw. The physical world is imprecise — it yields outcomes theory alone can't explain. We don't clean those deviations out. We capture them, quantify them, and let our models learn to design around them. That's the difference between simulating materials and making them.

Today @NVIDIAGTC we cemented our spot as the industry leader in the materials space. Autonomous labs are the future, and no one is building nearly as fast as our amazing team @RadicalAI.

Read here: radical-ai.com/news/beyond-co…

Add it to your agenda here: nvidia.com/gtc/session-ca…

If you're at NVIDIA GTC this week, join Radical Co-founder and CEO @josephfkrause's session on the AI-driven autonomous lab of the future: March 17 at 3 PT IRL and on livestream.

Computational chemists understand the physics. Software engineers understand distributed systems. They rarely work on the same problems in the same environment. That gap is why most materials simulations don't scale, and why scaled simulations often aren't scientifically defensible. Our team sits at the intersection. Chemists thinking about reproducibility and observability. Engineers thinking about what a scientifically meaningful test means. Our post on what happens when you bring both together to build structure-aware property prediction at scale.

Read here: radical-ai.com/news/beyond-co…

The most amazing part of Matt's work is the box full of nearly a hundred prototypes and versions that didn't work, before he was able to land on something that solved the problem. There's a lesson in every single one of those "failed" prototypes.

Apply to join here: luma.com/frontier-panel…

NYC still doesn't get the deeptech opportunity. It will soon. On March 31, @EricNewcomer will moderate a conversation with Radical CEO @josephfkrause and @faunarobotics' Rob Cochran at @southpkcommons NYC, organized by @ditzikow.

Read here: radical-ai.com/news/define-de…

Most engineers solve complex problems with complex solutions: more sensors, tighter tolerances, better control loops. We took the opposite approach. New post from our mechatronics engineer Matthew Borgatti on building robust automation by adding flexibility instead of complexity.

Read millions of publications simultaneously. Test billions of material candidates. Connect all of that information at once to form a hypothesis. The R&D process that has previously taken years is possible in days, and looks radically different driven by AI.

Apply to join here: luma.com/frontier-panel…

Deeptech isn't one category. It's robotics, materials, AI systems that do real work in the physical world. And it's happening in NYC. @josephfkrause and Rob Cochran @faunarobotics will talk about why this space is still under-hyped with moderator @EricNewcomer at @southpkcommons NYC on March 31. Big thanks to @ditzikow for organizing.