Joseph Krause

Joseph Krause is a materials scientist and co-founder of Radical AI, identified as its CEO in a September 2025 interview. His central founding decision was to combine AI discovery with robotic material synthesis and testing, with manufacturing and material sales as eventual goals. In a June 2026 Latent Space interview article, he defined discovery by whether a material reaches a useful product: “We count it as a discovery when you pick up your phone and there’s a new material sitting inside of it.”[2][1]

Krause described the founders’ choice as vertical integration: they wanted to build a materials company rather than stop at AI software or high-throughput experimentation. The intended connection was between generating hypotheses, making and testing materials, and using the resulting data to support further discovery and eventual scale-up. Manufacturing and selling materials at scale remained goals in his September 2025 account.[2]

He explained inverse design as starting with an application’s required property—such as a strong material for a jet turbine—and working backward to propose candidate materials. In the loop he described, AI hypotheses go to a robotic lab for synthesis and testing; analysis of those results informs which material to make next. He said Radical captures temperature, pressure, oxidation, concentration and even room humidity, feeding those experimental conditions back into its AI engine.[2]

Krause argues that experimental data supplies the industry’s competitive advantage because composition alone does not establish how a material will perform. As he put it in the June 2026 article, “In materials, the ground truth is the material itself. You have to be able to test it and characterize it.” He also proposed learning the mechanisms behind scientists’ decisions from experimental feedback as a route toward scientific intuition. That was an aspiration: in September 2025, he explicitly said, “We’re not there yet”.[1][2]

The June 2026 article reported that Radical had produced and characterized 1,200 alloys in six months. It separately attributed to Krause an estimate that the system could scale to 100 tested and characterized alloys a day. The first figure was reported output; the second was a prospective capacity estimate.[1]

Krause is a coauthor of the August 6, 2025 paper “TorchSim: An efficient atomistic simulation engine in PyTorch”, alongside Orion Cohen, Janosh Riebesell, Rhys Goodall, Adeesh Kolluru, Stefano Falletta, Jorge Colindres, Gerbrand Ceder and Abhijeet S Gangan. The paper lists Krause’s affiliation as Radical AI.[5]

Before founding Radical, Krause worked in semiconductor-materials research and venture investing, giving him experience on both the laboratory and company-building sides of commercialization. He earned a BS in Nanoscience from Virginia Tech and pursued graduate research at Rice University. Recalling that period, he said he wanted to commercialize the materials he was making in the lab. He subsequently held an Army Research Lab fellowship; AlleyCorp’s biography describes his research there as developing novel neuromorphic computing chips for Army applications and records six years of Army National Guard service.[2][3]

Krause recalled cold-emailing New York investors, including Kevin Ryan, to find someone from whom he could learn how companies are built. Ryan offered a six-month internship at AlleyCorp. Krause took a leave of absence and, by his account, spent about two and a half years at the firm before starting Radical. He worked on materials science and deep-tech investments, crediting Ryan with mentorship and room to make hands-on investment decisions.[2]

At AlleyCorp, Krause sat beside Jorge Colindres, whose interest in applying AI to difficult scientific problems prompted a conversation about materials science. Krause recalled that they spent the next two months reading hundreds of research papers together. They then teamed with Gerbrand Ceder, whose experience building autonomous laboratories Krause credited as part of the founding collaboration, and launched Radical AI.[2]

Krause publicly advocates rapid decisions followed by learning from results. In September 2025, he described Radical’s 51% rule: make a decision once confidence reaches that threshold, then iterate, because he considers delay more harmful than a wrong decision that produces learning. He also recalled temporarily banning the word “strategy” in an internal email to push execution. His explanation was practical: “the faster you can execute, the faster you can learn, and then the faster you can update your process from that learning.”[2]