Finding a molecule is the easy part. GaiaBT, a Korean startup founded in March 2024, argues the harder question is which patients a drug will actually help, and it has built a platform that tries to answer it with a loop of AI agents and wet-lab experiments.
The company runs two pieces of software. K-GEM is an operating system that links universities, hospitals, research institutes and biotech firms. GAIA-Phil is the engine on top, coordinating specialized agents that search data, run analyses and hand results to the next step.
No own lab, just partners and agents
GaiaBT has no internal laboratory. Partner institutions test its hypotheses in cell lines and organoids, the company folds those results back into its models, and the cycle repeats. Chief executive Kang Yunkyung describes design, experiment and verification as one continuous structure.
Beyond genomic data, the platform reads epigenomic signals, which regulate gene expression without changing the DNA sequence, as clues to drug response.
Its lead candidate, Onco101, targets cholangiocarcinoma. GaiaBT wants to predict which patient subgroups respond and sharpen selection criteria using biomarkers. It also signed a 50 million won contract with Cyron Therapeutics to optimize an antibody-drug conjugate platform.
The footprint is wider than the headcount. GaiaBT has five employees, who use agents for grant applications and investor materials as well as research. It holds a joint research contract with Cedars-Sinai and works with Asan Medical Center and KRIBB, with MOUs at 13 institutions.
Plans call for an agent orchestrator in 2027, Onco101 entering preclinical work the same year, US expansion, and an IPO target of 2030.
