The AlphaFold judge who built an AI drug company

"The world stars are Psy — and Seok Chaok."

That line, delivered on a public broadcaster's science program, has been making waves on YouTube. The subject is Seok Chaok, a chemistry professor at Seoul National University and co-CEO of Galux, an AI-driven drug development venture.

Seok has spent more than two decades pioneering computational chemistry and currently serves as head of the science subcommittee of the Presidential Committee on Artificial Intelligence, shaping national AI science policy. A YouTube video of his appearance on EBS's "Hobby Is Science," in which he discusses protein structure and AI innovation, surpassed 390,000 views within two days of its release, drawing widespread praise online.

Much of the public's fascination centers on a behind-the-scenes account of the moment Google DeepMind's AlphaFold 2 shook the scientific world in 2020. Seok had for years been a top-ranked competitor at CASP (Critical Assessment of Structure Prediction), the biennial global protein-structure prediction competition, and was serving as a judge when AlphaFold 2 made its landmark appearance.

Recalling the moment, Seok said he felt "both profound shock and deep joy" when he saw scores he had believed impossible in his lifetime. He also vividly recounted how, during the online results announcement, DeepMind researchers fielded a barrage of questions from scientists around the world but left without disclosing their algorithm — saying only that they had solved it with AI and were hiring talent.

Seok noted, however, that the technology exemplified by AlphaFold remains at the "prediction" stage — identifying the three-dimensional shape of existing proteins. While it has undeniably spared researchers decades of painstaking experimental work, prediction alone does not readily translate into developing treatments that cure disease.

Galux, which Seok co-founded with his students, has pushed that prediction capability into a "design" stage. The company's core technology is de novo antibody design — using computational methods to engineer entirely new antibodies that do not exist in nature, built from the outset rather than discovered by chance. Where conventional approaches rely on animal immunization or large-scale library screening and can take more than a year just to identify a candidate, Galux uses its proprietary AI platform GaluxDesign to engineer antibodies that bind to a chosen disease target, cutting the discovery window to under a month.

Galux combines physicochemical principles with AI to achieve atomic-level precision — within a margin of one angstrom. The company claims world firsts in de novo antibody design against six therapeutic targets, including the structurally uncharacterized ALK7 as well as PD-L1 and HER2, and has demonstrated the approach's scalability to high-difficulty targets such as GPCRs and ion channels. Using just 50 refined designs per target rather than broad screening, Galux has recorded a binding-antibody success rate above 30 percent — thousands of times more efficient than conventional methods, the company says.

The technology has also held up against global competition. In a comparative study pitting Galux against leading international AI platforms — including Nabla Bio's JAM-2 and Chai Discovery's Chai-2 — across nine shared targets, Galux secured binding antibodies for eight, outperforming Nabla Bio (five targets) and Chai Discovery (four) to claim the top success rate.

That track record earned Galux a spot on Forbes Asia's 100 to Watch list for 2026, making it the only South Korean biotech or healthcare company on this year's roster. The company is collaborating with global pharmaceutical giants AstraZeneca and Boehringer Ingelheim, and its independently developed PD-1/IL-18v bispecific immuno-oncology antibody has been selected for support under the national new drug development program. Galux has raised a cumulative 70 billion won ($50.6 million) in investment as it accelerates its push to become a next-generation global biotech.

The YouTube video has drawn an outpouring of supportive comments, with viewers writing that de novo protein design — engineering new proteins with desired functions using AI — is "fascinating," that Seok is "incredibly cool," and that they "sincerely hope a Nobel Prize is not far off."