**Author**: Will Thompson (Twitter)

Some notes on what you might expect during a quant interview - from someone who has been through a few.

- Intro
- Disclaimers
- The Quant-Dev Spectrum
- A Survey of Types of Interview Problems
- Maths Questions
- ML Questions
- Coding Questions
- *Footnote
- Citation

# Intro

Let’s say you are a SWE that stumbled upon the online community Blind. You’ve likely been inundated with multiple “prestige rankings” posts.

What stands out is that “quant finance” firms seem to always find themselves in the highest rungs.

Intrigued, you decide to apply (likely through a headhunter) for a few “quant” positions (quant researcher, quant dev, research engineer, etc)** ***.

They bite and now you are scheduled for a few interviews. You got your foot in the door 🚪 - now you have to figure out how to prepare.

*Disclaimers*

I haven’t sat through a quant interview in a long time. **It’s likely that the expected value of this advice is ~ 0**. And unlike FAANG interviews, there is a** high degree of variability** in the types of questions you might be asked.

This is meant as a starting point for those looking to prepare.

# The Quant-Dev Spectrum

You often see a number of titles across companies: Quant Trader, Quant Researcher, Quant Analyst, Quant Developer, Research Engineer, etc etc. The reality is that there is no objective ground truth; **the meanings behind these titles vary widely across firms and even more so across types of quant firms**. Ultimately, it depends on the type of systems these firms deploy to garner an edge in the market.

For instance, the role could entail tweaking the trading logic of a live trading system. It could mean searching for orthogonal predictive signals within a set of noisy financial data. It could mean writing counterfactual exchange simulation code and debugging low-latency production systems. Again, it depends on the type of system.

In fact, multiple roles may hold the same title within a firm. To assess where you might fit, these firms are trying to place you on what some call the “**quant-dev**” spectrum.

# A Survey of Types of Interview Problems

*Here’s a survey of the type of questions you might encounter:*

## Maths Questions

__Note__: These questions are more likely to be asked if you will be doing research or strategy development.

**Probability**: Expect some probability and expectation-based questions. Questions surrounding**flipping****coins**,**dice**, and**urns**are very popular (as they were if you took college-level probability theory). Some of the questions can be, but are not limited to:**Calculating the number of rolls required to get 2 6’s with a single die**- Calculating the expected value of a re-roll with dice
**How to simulate dice using coin flips**(i.e. rejection sampling)- Calculating the upper and lower bounds on the correlation between 2 linked random variables
- The expected outcome of some kind of repeated games where an individual players has probability
*p*of winning - Some coin flip problem about ascertaining the probability that it’s an unfair coin (i.e.
**Bayes’ Rule**) **Linear Algebra (less often)**: They*might*ask questions surrounding PCA, calculating covariances, diagonalizing a matrix, or why a singular matrix is problematic, etc.

**Personally, I think the book ****50 Challenging Problems in Probability**** is a good survey of the types of probability questions you should expect to know**. It’s a pretty small book.

## ML Questions

Financial data is often **noisy**, **multi-regime**, and **non-stationary** with **concept drift **(and usually not enough N). Often these questions deal with how you handle modeling in that type of environment, particularly if you have a** large collection of correlated weak learners.**

**OLS**: This is the workhorse model of most prediction problems in finance. You are expected to understand the properties of OLS fairly well and in particular be able to discuss in-depth**OLS**versus say**Lasso**and**Ridge**Regression.**Time-Series**: Financial data tends to also be non-stationary. You might be asked how you might detect a unit-root or re-frame the problem to model a stationary process (particularly if your interviewers love econometrics**,**i.e. U Chicago Econ PhDs).**This is more likely in a longer-horizon fund.**- Usually, deep learning is not discussed, although, it’s been a long time and this might have changed!

**Back in the day, ****ESL**** was a good book to skim through.**

## Coding Questions

These can be fairly straight-foward leetcode style questions. Sometimes, particularly if a system relies heavily on a certain language, they might assess your competency of that language.

In my experience, these usually surround some form of computational problem. For instance, they may ask you to implement your **favorite ****shuffling algorithm** or to implement and discuss **the properties of a ****reservoir sampling algorithm** or different ways of attacking the median of medians problem.

# *Footnote

There are reasons other than external validation to check out quant finance - research that is immediately impactful to the core business; pushing code frequently; motivated peers; ownership/autonomy; “intellectual fulfillment”; in some situations, low-level system design/ latency optimization; etc etc.

Like any job, though, there are cons to consider as well (i.e. “no free lunch”). **Here’s one ex-quant’s ****take**.

# Citation

```
@article{
title = "Grokking a “Quant” Interview",
author = "Thompson, Will",
journal = "https://willthompson.name",
year = "2024",
month = "08",
day = "23",
url = "https://willthompson.name/grok-quant-interview"
}
```