Description
UbiquiTx builds genetically encoded therapeutics that edit the proteome. Our core platform fuses short, designed peptide binders to post-translational modification enzymes (E3 ligases, deubiquitinases, kinases, phosphatases) to selectively modulate disease-driving proteins in cells. All binders are designed in-house using our own generative models and validated experimentally. We are hiring a computational scientist to work directly on model execution, data generation, and learning systems that close the loop between experiments and model updates.
What you will do (near term):
Run and adapt our existing peptide and protein generative models for new targets.
Design and launch binder libraries using our internal workflows.
Manage inference, sampling, filtering, and ranking of candidates.
Work closely with experimentalists to translate design outputs into testable constructs.
Debug model behavior, sampling pathologies, and failure modes in real design campaigns.
What you will build (medium term):
Systems to learn from experimental outcomes and feed that data back into model training.
Data pipelines that connect binding, degradation, or PTM readouts to model weights.
Training and fine-tuning strategies that exploit sparse, noisy biological data.
Evaluation and tracking tools to understand when and why models succeed or fail.
Infrastructure that supports iterative design rather than one-off runs.
Requirements
Background we’re looking for:
Strong Python and PyTorch experience.
Hands-on experience running large neural models, not just prototyping them.
Familiarity with protein or peptide modeling, protein language models, or generative models.
Comfort working with imperfect data and tight experimental feedback loops.
Interest in building systems that actually get used by wet-lab teams.
Nice to have:
Experience with diffusion, flow matching, or other modern generative methods.
Experience designing or analyzing biological datasets.
Exposure to model retraining from experimental or real-world feedback.