What the work actually looks like

You are handed a research-grade analysis task — often with raw or deliberately messy data — and asked to carry it through end to end: inspect the dataset, run quality control, do exploratory analysis, choose and justify statistical methods, execute the analysis in R or Python, and write up the interpretation. Tasks are drawn from computational genomics, quantitative biology and translational biomedicine: bulk and single-cell RNA-seq, variant calling output, GWAS or QTL summary statistics, population-genetics panels, multi-omic integrations. Some assignments ask you to produce the reference analysis yourself; others ask you to review a model's attempt and mark where its QC was skipped, its test was misapplied, or its conclusion outran the data.

The distinguishing feature is ambiguity. Prompts rarely tell you which normalisation to use or which covariates matter. Part of what is being captured is your decision trail — why you dropped those samples, why you chose a negative binomial over a t-test, which confounder you checked for and how, and what the analysis genuinely cannot resolve given the data at hand.

What the screen is measuring

  • Bench-level fluency, not vocabulary. Expect follow-ups on specific tools, parameter choices, and what a diagnostic plot looked like when something went wrong.
  • Statistical judgment under imperfect data. Batch effects, multiple testing, confounding by ancestry or cell composition, power limits.
  • Code you can defend. You should be able to write, debug and explain analysis code in both R and Python, not just one.
  • Writing. Methods, assumptions and uncertainty stated precisely — graders read your prose as closely as your code.

Logistics

Fully remote and asynchronous, open globally. Engagements are project-based: task batches arrive with deadlines rather than fixed shifts, and most contributors work somewhere between 10 and 20 hours a week. A single substantial analysis task can take several hours, so contiguous blocks of focused time matter more than daily availability. Pay is observed around $120/hr and varies with domain depth and task complexity; it is not guaranteed. Onboarding typically includes a paid or unpaid calibration task before regular volume begins.