Research Scientist

The problem.

Everything the firm believes rests on a small set of statistical decisions: how windows are split, what an embargo protects against, how many effective trials a search ran, what a confidence interval means under dependence and non-identical distributions. Those decisions are currently made well by people who are not primarily statisticians. You would make them better and write them down.

What you would do.

• Take the firm's validation machinery apart. Say where it is defensible, where it is conventional, and where it is wrong.
• Build the estimators the desk needs and cannot correctly implement in an afternoon.
• Quantify the search: effective number of trials, dependence between configurations, and what survives adjustment for both.
• Write internal notes that make a method usable by someone who did not derive it, with assumptions stated where they can be checked.
• Publish externally, at the firm's quarterly cadence, about method only — never about positions.

What we would want evidence of.

• You can derive a result and say what breaks when its assumptions fail.
• You have worked with dependent, non-stationary observations and did not reach for independent-samples formulae out of habit.
• You can tell a rigorous method from an elaborate one, and prefer the first.
• You have written something a practitioner actually used.
• You are willing to be the person who says the study is underpowered.

What would make this a bad fit.

This role is a poor fit if you want a research agenda insulated from production. Every method you build is used within weeks and questioned shortly after.

This role is a poor fit if publishing is the point of the work. We write about method and never about what we trade. That rules out much of what a paper would normally contain.