You will create rigorous research tasks for AI training and evaluation using real-world biological datasets. Each task should lead to one objectively verifiable answer and include clear scientific reasoning, evidence, and reproducible documentation.
- Design multi-step tasks using sequencing, genomics, transcriptomics, proteomics, metabolomics, structural biology, and related datasets.
- Source datasets from authorized repositories, follow licensing requirements, and remove identifiable patient information.
- Create reproducible Python validation notebooks that show how answers are derived from the underlying data.
- Establish ground truths using PCR-confirmed results, known gene knockouts, verified pedigrees, or other independently validated biological information.
- Sanitize file names, headers, and metadata so task materials do not reveal answers.
- Separate objectively verifiable biological facts from conclusions that depend on interpretation.
- Document workflows accurately and work with research and technical stakeholders to improve task quality.