Small team, unusually short path from idea to experiment
You will work on problems where the computational and experimental sides touch directly. Models you train this month get tested at the bench, and the result comes back to you rather than to a report.
Daejeon based
On site with our collaborators, with flexible hours and remote days. Research in a loop needs people who can walk to the instrument.
Publication encouraged
Methods work can be published, including negative results, subject to partner confidentiality.
Equity for everyone
Every employee holds equity. Compensation is discussed openly in the first conversation, not at the offer.
Real compute
Training and inference budget that matches the problem, plus a standing allowance for conferences and courses.
Five positions
If none of these fits but you think you should be here, write to us anyway and say what you would work on.
Machine Learning Scientist, Nucleic Acid Models
Machine Learning Scientist, Nucleic Acid Models
You will own the generative models at the centre of the design loop: training and adapting sequence models for RNA and DNA, conditioning generation on structural and developability constraints, and making the uncertainty of a prediction as useful as the prediction itself.
What you would do
- Adapt pretrained nucleic acid language models to aptamer design tasks
- Build conditional generation that respects motif, thermodynamic and composition constraints
- Design active learning strategies that choose which candidates to synthesize
- Calibrate model confidence against assay outcomes, round after round
- Work directly with the chemistry and assay teams on what is measurable
What we look for
- Strong background in deep learning for sequences, with published or shipped work
- Comfort with biological data that is scarce, noisy and expensive
- Interest in calibration and uncertainty, not only benchmark accuracy
- Willingness to have a model's output falsified by an instrument
Computational Structural Biologist
Computational Structural Biologist
You will own how we reason about structure: folding ensembles, tertiary prediction, docking and the scoring functions that decide which candidates reach synthesis. This role sets the quality ceiling of everything downstream.
What you would do
- Build and validate secondary and tertiary structure pipelines for designed candidates
- Develop scoring that accounts for ensembles and ionic conditions, not single structures
- Model protein binding surfaces and run docking against designed binders
- Quantify where our structural predictions are unreliable and make that visible
- Design the controls that distinguish a folding failure from a binding failure
What we look for
- Deep familiarity with RNA secondary structure methods and their limits
- Experience with molecular docking or molecular dynamics
- Scientific scepticism about prediction accuracy, backed by experience
Oligonucleotide Chemist
Oligonucleotide Chemist
You will decide what is actually makeable. Modification strategy, synthesis feasibility and the chemistry constraints that enter the design specification before a single candidate is generated.
What you would do
- Define modification strategies for nuclease resistance and half-life extension
- Encode synthesis feasibility and cost into design constraints
- Oversee synthesis and purification of shortlisted candidates with our collaborators
- Troubleshoot cases where a modification destroys the intended fold
- Develop conjugation chemistry for payload and polymer attachment
What we look for
- Hands-on solid-phase oligonucleotide synthesis experience
- Knowledge of modified nucleoside chemistry and its structural consequences
- Analytical skill with liquid chromatography and mass spectrometry
Biophysics & Assay Development Scientist
Biophysics & Assay Development Scientist
You will make the measurements trustworthy. The loop learns from your data, so artifacts do not merely waste an experiment, they corrupt the next design cycle.
What you would do
- Develop and run surface plasmon resonance and biolayer interferometry assays
- Design controls that expose mass transport, avidity and nonspecific binding
- Standardize folding and buffer protocols so results are comparable across rounds
- Build the structured data handover that feeds post-assay analysis
- Push back on design proposals that cannot be measured cleanly
What we look for
- Practical experience with label-free binding kinetics instruments
- Instinct for distinguishing a real signal from a surface artifact
- Care about data provenance and reproducibility
Software Engineer, Agent Infrastructure
Software Engineer, Agent Infrastructure
You will build the system the scientific agents run on: the tool interfaces, the data layer that keeps every candidate traceable, and the audit trail that makes an agent's decision reviewable by a scientist.
What you would do
- Build reliable tool interfaces between agents and scientific software
- Design the data model linking candidates, constraints, scores and assay results
- Make the reasoning trail queryable so reviews can be substantive
- Orchestrate compute for generation and scoring workloads
- Instrument the loop so we can tell which changes actually improved outcomes
What we look for
- Strong backend engineering, with Python and workflow orchestration experience
- Experience building systems around machine learning rather than only models
- Interest in scientific tooling and in correctness over throughput
What applying looks like
Email us
Send a CV and a few paragraphs on a problem you have worked on and what you got wrong. We read every application ourselves.
Technical conversation
A discussion of your past work in depth, and of a real problem from our current loop. Compensation range is covered here.
Work session
A paid, time-boxed exercise close to the actual job. No puzzles, no unpaid take-home projects.
Decision
Meet the team, then a decision within a week either way, with specific feedback if the answer is no.