The work
You will design scientific workflows that test whether AI agents can complete authentic, multi-step life sciences analyses. Each workflow should require agents to inspect data, choose suitable methods, run code, troubleshoot results, and produce scientifically valid conclusions.
You will create the complete task package and check that it works in a controlled, network-isolated environment.
- Develop scientific datasets, input files, instructions, constraints, and expected deliverables.
- Write expert solutions and create objectively verifiable ground truths.
- Build automated or semi-automated grading criteria for evaluating agent work.
- Validate scientific assumptions, calculations, code, intermediate outputs, and final answers.
- Design workflows that are self-contained, reproducible, and executable without network access.
What it pays and takes
This is a remote contractor assignment lasting five weeks. The work requires advanced life sciences research experience and strong scientific programming skills, especially in Python.
- Pay: $300 per approved task.
- Schedule: At least 6 hours per day and 40 hours per week.
- Time zone: At least 4 hours of overlap with Pacific Time.
- Language: English.
- Experience: Ph.D., postdoctoral, or equivalent research experience in the life sciences.
- Required skills: Multi-step computational scientific analysis, Python, and independent validation of scientific reasoning and computational results.
- Relevant backgrounds include bioinformatics, computational genomics, systems biology, computational neuroscience, biostatistics, computational drug discovery, computational biochemistry, structural biology, protein engineering, and computational microbiology.
How it works
Apply on OpenTrain with your resume, then complete the application on the hiring site.
About AI training work
AI training is the human work behind systems that learn from examples and feedback. In this role, your scientific expertise helps create reliable tasks and evaluations for AI systems that must perform real computational research workflows.