A design company that happens to work on medicine
Sanghyun was founded on a specific frustration: the tools to design nucleic acid binders computationally had become good enough, while discovery was still being run as a selection experiment from 1990. We built the engine we wanted to use.
Founded in Daejeon, Republic of Korea, alongside the research institutions and wet labs we collaborate with.
No clinical programmes and no claimed results. We are building the engine and testing it against our first targets.
Deliberately placed in Korea's research corridor, where the instruments and the expertise to run the loop are a short walk away.
Make designing a binder as ordinary as designing a circuit
An engineer designing a circuit does not breed a million circuits and keep the ones that conduct. They model, simulate, build the shortlist and measure, and the measurement improves the model.
Binder discovery still mostly works the other way around. Selection is a remarkable technique, and it remains the right tool for some problems, but it is a search strategy that cannot explain its own results. When a campaign fails, it rarely tells you why.
Our bet is that the combination of specialized nucleic acid models and agentic scientific reasoning makes design tractable. Not perfect, and not a replacement for the bench, but tractable enough that the bench becomes the place where hypotheses are tested rather than where candidates are found by chance.
Who runs this
Founder · Chief Executive Officer & Chief Technology Officer
Sanghyun Park
Sanghyun Park founded the company and holds both the chief executive and chief technology roles, which in a company at this stage is less a title than an accurate description of the work. Sanghyun sets the scientific direction of the engine, owns the architecture of the design loop, and leads partnering conversations directly rather than through a business development layer.
The founding thesis came from working at the boundary between computational modeling and experimental validation, and from watching how much information a conventional campaign throws away. Every candidate that fails to bind is a measurement. Treating those measurements as training data rather than waste is the reason the engine is built as a loop.
Sanghyun is based in Daejeon and is the right person to contact about technical collaborations, partnering structures or the design methodology itself.
Five commitments
These are operating rules rather than values on a wall. Each one costs us something, which is how you can tell they are real.
Report what the data says
Including when it contradicts the design hypothesis, the partner's expectation or the previous quarter's conclusion. A platform whose outputs are always encouraging is not measuring anything.
Keep the reasoning inspectable
Every shortlisted candidate carries the constraints it met and the scores behind it. If a reviewer cannot disagree with a specific step, the audit trail has failed.
Scientists decide, agents prepare
Automation handles the data, the scoring and the analysis. A person approves the specification and the synthesis list. We are not interested in removing the judgment.
Publish the methods, not just the wins
We write about where our approach falls short, because a field that only publishes successes takes longer to find out what works.
Say no to targets we cannot serve
A scoped collaboration that ends in a clear negative is more useful to a partner than an open-ended one that never resolves.
Treat the wet lab as a peer
Computational design without experimental partners is literature. Our collaborators shape the design questions, not just answer them.
The loop needs a bench
We work with wet labs at KAIST for synthesis and high-throughput binding characterization, including surface plasmon resonance, biolayer interferometry and deep sequencing of selection pools.
That relationship is central rather than incidental. It sets how fast we can learn, and it keeps the engine honest: a design that only looks good in silico gets found out within a cycle.
Capabilities the loop needs
- Solid-phase oligonucleotide synthesis with modified chemistries
- Surface plasmon resonance for association and dissociation kinetics
- Biolayer interferometry for parallel screening
- Next-generation sequencing of selection pools
- Cell-based functional and internalization assays
- Nuclease stability and matrix stability testing
Come build the engine
We are hiring across machine learning, nucleic acid chemistry and assay development.