What the work involves

You spend most of your time reading quantitative analysis produced by a model and deciding whether it is actually correct — not whether it sounds correct. That means checking a derivation of an option price for sign errors and boundary conditions, catching a covariance matrix that was estimated on overlapping returns and treated as independent, noticing when a backtest silently uses point-in-time data it could not have had. The rest of your time goes to authored demonstrations: you write the reference solution a model should have produced, including the code, the assumptions you made explicit, and the parts of the problem where a defensible answer requires stating what you cannot know.

Tasks arrive in batches and typically cover derivatives pricing, portfolio construction and optimization, risk measurement (VaR, ES, stress design), time-series and factor modeling, and execution or microstructure questions. Some batches ask for pairwise ranking of two model responses; others ask for a rubric-scored critique with written justification. Rationale quality is the deliverable as much as the label — a rank with no stated reason is generally rejected.

What the platform screens for

  • Depth that survives a follow-up. The screen will take a topic you named and push one or two levels past the textbook statement. Naming a method is cheap; explaining when it fails is the signal.
  • Working code fluency. Python, R, or C++ used for real quantitative work — not reading familiarity. Expect to describe what you have actually built and run.
  • Calibration on ambiguity. Many finance questions have no single right answer. Reviewers who declare a defensible answer wrong because it differs from their house style score badly.
  • Written reasoning. Async work means your justification is the entire artifact. Terse, structured, specific writing is measured.

Logistics

Fully remote, fully async, no standing meetings. Contractors commonly commit 10–20 hours a week and set their own hours, though throughput expectations exist per batch and unclaimed work returns to the pool. Pay is hourly and reported in the $150–200 range for this specialty; rates are observed, not guaranteed, and vary with task complexity and calibration performance. Onboarding usually includes a paid or unpaid calibration set before full task access opens.