Iterative Labs
Example engagement

Environment designer, protein and antibody design

Iterative Labs builds reinforcement learning (RL) environments that teach frontier AI models to do science. An agent ranks protein or antibody candidates in simulation, and the batch it picks is measured on a real assay that returns a scalar within 48 hours. We need a protein engineer to turn design campaigns into those environments: choose the task, build the assay oracle, freeze the baseline the agent must beat, and specify the plate that settles it. One environment is one task, and the first is ready now.

What you will do

  • Select protein and antibody design tasks that fit our criteria (a ranking output, a scalar instrument readout, a bench round under 48 hours) and score them against our task criteria.
  • Build each environment as a Harbor task: the problem statement, the assay sidecar the agent queries for batched measurements, reference solutions and the automated verifier.
  • Freeze the in-silico baseline, define the reward as lift over it (precision@k on the agent's batch), and write the ablations that show the task is neither trivial nor unsolvable.
  • Specify the real-world check: the assay, the plate layout, the readout that becomes the reward, a cost a contract lab can quote, and the biosecurity screen the target passes.
  • Set each pass threshold from our own measured replicate standard deviation, not a published one.

You should have

  • Research experience of your own: a paper, preprint or thesis chapter you authored on protein or antibody design, directed evolution, enzyme engineering or a comparable campaign.
  • Current PhD or master's student, postdoc, or a graduate within the last two years, in protein engineering, biochemistry, structural or computational biology, or bioengineering.
  • Daily use of AI coding and agent tools, and a specific account of where they fail you on scientific work and how you check their output.
  • Working use of at least one protein design or structure-prediction tool (ProteinMPNN, RFdiffusion, BindCraft, AlphaFold, Boltz, Rosetta), shown by a repo or notebook.
  • Comfortable writing Python in a terminal and in Docker containers.

Nice to have

  • Hands-on wet-lab experience with a protein readout you could hand to someone else as a protocol: expression and purification, cell-free expression, a plate-reader assay, thermal shift, or binding by biolayer interferometry (BLI) or surface plasmon resonance (SPR).
  • Designs of yours that someone else synthesised and measured, and how many worked.
  • Batch selection under a measurement budget: active learning or Bayesian optimisation.
  • Antibody-specific work: nanobody or VHH scaffolds, humanisation, or developability screening.
  • Time running or operating an automated lab or biofoundry platform.

How we work

  • A hard scope line: benign, defined, non-pathogenic targets, industrial, diagnostic and basic-research endpoints only, cell-free and in vitro. No pathogens, toxins, transmissibility, human subjects or animal work.
  • Our cell-free biology station is planned and not yet commissioned. Until it is, bench rounds run at contract labs, and you help specify what it must measure.
  • The environments team is small; you work directly with the head of environments and the engineers who maintain the Harbor tooling.

Compensation and logistics

  • Compensation: USD 1,500 per task, each task about 10 to 20 hours of work.
  • Employment type: contract, paid per task, with more tasks as the catalogue grows.
  • Location: remote, Canada or the United States.
  • Start date: immediate.

Iterative Labs considers every applicant on the evidence above, without regard to any protected characteristic.

Join the expert network