Squarepoint Capital is a global systematic investment manager that spun out of Barclays' quantitative trading business in 2014, running data-driven strategies from offices across Europe, the US and Asia. Its researcher interview is dominated by one stage candidates describe in detail: a superday built around a six-hour data exercise, ending with an hour presenting your model to a researcher. Everything before it is a filter; that day is the evaluation.
The superday is a compressed research project: dataset in, analysis and predictive model out, then an hour defending it. No other stage matters as much, and no stage rewards validation discipline more directly.
- 1HR screen. Background and motivation.
- 2Technical round one. Reported as two medium-level coding problems plus one probability question.
- 3Technical round two. Statistics: linear regression and hypothesis testing.
- 4Superday. A roughly six-hour data exercise (one reported instance involved a credit dataset), an hour-long presentation to a researcher, and a finance-focused interview with a senior researcher.
Stages 1 to 3: The filters
The early technical rounds are conventional and preparable: medium-level algorithmic coding, a probability question, then a statistics round on regression and hypothesis testing done properly rather than recited.
How to prepare
Drill mediums until routine, keep the brainteaser bank warm for the probability layer, and make sure regression is something you can implement and interrogate, not just call. The first two problems below carry this firm's tag: one is a directly reported task, the other was built from the validation territory the data exercise grades:
Stage 4: The six-hour data exercise
You are handed a dataset and asked to analyse it and construct a predictive model, then present the result for an hour. Under time pressure, the instinct is to maximise the model; the winning move is to budget the day like a researcher: understand and clean the data, build something defensible, and reserve genuine time for validation and the story you will tell.
The presentation hour is the project-defence round with the ink still wet: expect why-this-model, what-did-you-check, what-would-break-it. A candidate who says "here is where it leaks, here is what I would do with a week" beats one who claims six-hour perfection.
How to prepare
Simulate the day once before the real one: take any public dataset, give yourself six hours, and end with a ten-minute writeup. Use the validation course as the checklist for the middle hours, and its twenty questions as the rehearsal for the presentation.
Key tips for success
- Practise working fast on unfamiliar tabular data: loading, cleaning, summarising and joining without friction is what buys you modelling time.
- Decide your time budget before the superday starts, and protect the final hour for validation and the writeup.
- In the presentation, lead with the honest caveats; the interviewer is a researcher, and researchers grade scepticism.
Closing remarks
Squarepoint runs the most literal version of the project-driven interview: they give you data and watch you research. The filters ahead of it are standard; the day itself is won in preparation, by having done a time-boxed research exercise before, so the real one is a repeat performance.