Overview
We’re looking for experienced bench scientists to help train AI models. You’ll design realistic experimental-biology tasks from your own practice, run them through frontier AI systems, and grade what comes back against the standard you’d hold a colleague to.
The models can already talk fluently about biology. What they can’t yet do reliably is the real work: call a plate QC go/no-go, gate a flow panel, check a cloning strategy, or read a Western blot from the raw images. Your bench judgment becomes the benchmark those models are measured against.
What you’ll actually do
- Design realistic experimental tasks from your own workflows: the scenario, the prompt, and the supporting files an agent would need (plate maps and reader output, gel and blot images, flow summaries, protocols, QC records).
- Run tasks through frontier AI agents and grade the deliverable they produce (a QC summary, experimental plan, or troubleshooting memo) against a professional standard.
- Write clear grading rubrics that specify what a correct deliverable must contain (the right controls, the right wells excluded, the right diagnosis of a failed run), and explain why a response passes or fails.
- Flag concrete failures with evidence: controls ignored, edge effects and outliers missed, wrong normalization, plausible-sounding troubleshooting the raw data contradict, or fabricated values.
Problems draw on whatever you know best
- Molecular and cell biology: cloning, cell culture, and cell-based assays.
- Protein biochemistry: purification, characterization, and assay development.
- Immunology and flow cytometry: panel design and gating.
- Imaging and histology, microbiology and virology.
- In vivo and preclinical work, plus lab operations and analytical sciences.