RepostedAkuna Capital logo

Akuna Capital · Phone screen

In process

Akuna Capital QR Probability Phones: 14 Experiences (Part 5 of 6)

Quantitative researcher · UnknownReported Feb 28, 2019
Shared byjanebeingjane

Akuna Capital quant researcher interviews from 2018–2019: 14 reports with uniform sums, circle-arc proofs, regularization and SVMs.

At a glance

  • Process (2018–2019): usually one or two technical phones mixing probability, linear algebra and ML (H54–H67).
  • Asked in these reports: probability in 11 of 14 and ML in 8; repeats include uniform-sum questions (H55, H58, H59, H62, H63) and circle-arc problems (H56, H57, H59).
  • Difficulty: thresholds in the uniform-sum question changed between candidates (H63).

Coverage: 14 Akuna Capital quantitative research reports from March 2018 to February 2019 (QR intern, QR, Junior QR, JQR-ML). All are more than three years old, so they sit in History below.

In this series

  • Part 1: 2022–2025 screens, coding and the newest reports.
  • Part 2: 2021–2022: coding, simulation and probability.
  • Part 3: 2020: root finding, regression and shuffles.
  • Part 4: 2019–2020: ML and final rounds.
  • Part 5 (current): 2018–2019: probability and SVMs.
  • Part 6: 2017–2018: JQR, ML and projects.

Newer reports: Akuna quant technical interviews, 2025–2026.

What came up

The same probability families kept returning: sums of uniform variables, arcs on a circle and waiting times for patterns, with ML questions on regularization, k-means and SVMs. Every question stays in its entry.

Common questions

Which Akuna probability questions repeated in 2018–2019?

Sums of uniform variables (H55, H58, H59, H62, H63) and random arcs on a circle (H56, H57, H59).

What ML did Akuna ask junior quant researchers?

Regularization (H54), k-means (H57), SVMs and optimization (H64, H65), and regression and PCA (H67).

History (before 2024)

H54: Waiting for HHT and explaining regularization

Report date: 2019-02-28 · Role: Quantitative Research Intern

The phone invitation arrived roughly two weeks after the earlier assessment. More than ten minutes went into projects before two technical questions.

Technical phone

  1. A coin has P(heads)=p. Find the expected number of tosses until HHT appears: two heads followed by a tail.
  2. Explain L1 and L2 regularization, their effects and why they work.

H55: Uniform sums, matrix equations and regression variance

Report date: 2019-02-13 · Role: Quantitative Researcher

This account describes the mathematics part of a phone. Coding was described as a separate second round.

Technical phone

  1. Find the distribution of X+Y for independent U[0,1] variables.
  2. Find all square matrices A satisfying A²=I.
  3. If A^n=I, is A diagonalizable? The field and restrictions on n were not stated.
  4. Add predictor X2 to a regression that originally used X1. Does the variance of the estimated X1 coefficient increase or decrease?

H56: A neural-network project and a circle-arc proof

Report date: 2019-02-11 · Role: Quantitative Researcher

The phone covered the candidate's background, a neural-network project and a probability proof.

Technical phone

  1. The interviewer asked about a neural-network project on the candidate's resume.
  2. Cut a circle at three random points and find the expected length of the arc containing (1,0). After a symmetry argument, the interviewer wanted conditional-probability and integration reasoning. The candidate discussed deriving a CDF, then a density and expectation.

H57: JQR: K-means, circle arcs and containment probabilities

Report date: 2019-02-08 · Role: Junior Quantitative Researcher

The JQR phone opened with background and a data-related project. The candidate thought they had failed.

Technical phone

  1. What is K-means clustering?
  2. Cut the unit circle at three random points and find the expected length of the arc containing (0,1).
  3. What is the probability that a triangle formed by three uniform points on a circle contains the origin?
  4. Extend the question to three dimensions: what is the probability that four random points on the unit sphere form a tetrahedron containing the origin?

H58: Research, an ML project and part of a sum distribution

Report date: 2019-01-06 · Role: Quantitative Researcher; seniority not stated

Technical phone

  1. Describe the candidate's research.
  2. Discuss an ML project and its technical details.
  3. For a sum of independent U(0,1) variables, derive the distribution on the part where the sum is below 1. The candidate explicitly narrowed the question to that portion.

H59: Uniform sums and an arc-length generalization

Report date: 2018-12-31 · Role: Quantitative Research Intern

The interviewer arrived ten minutes late. Resume, internship and project discussion and two probability questions took a little more than 30 minutes.

Technical phone

  1. For N independent U(0,1) variables, find P(sum < 1).
  2. Cut the unit circle at three random points and find the expected length of the arc containing (1,0). Then extend it to n cuts.

H60: Two math/coding phone pairs

Report date: 2018-12-19 · Role: Quantitative Researcher, ML; seniority not stated

Both the first technical phone and the final phone contained mathematics and coding. The candidate reported failing the final phone.

First mathematics phone

  1. Derive the distribution of the sum of n independent U(0,1) variables, starting with two and using induction.
  2. Find the possible eigenvalues of an n×n matrix with A^n=I, as the candidate remembered it.
  3. What happens to coefficient variance when another predictor is added to a one-predictor linear regression?

First coding phone

  1. Why would you make a destructor virtual?
  2. How could you avoid writing a destructor repeatedly in a class? The candidate was not certain of the original wording.
  3. Write a recursive function that prints a formatted string. Neither the format nor the input was retained.

Final mathematics phone

  1. With 100 fair coin flips, what is the probability of an even number of heads?
  2. Cut the unit circle at three random points and find the expected length of the arc containing (1,0).
  3. Write a program to find an nth root. The candidate did not report a required numerical method.

Final coding phone

  1. Implement a maximum tracker. The account did not specify the operations.
  2. An ML project on the resume came up in discussion.

H61: An early-December phone reported later

Report date: 2019-02-09 · Interview period: early December 2018 · Role: Quantitative Research Intern

The first phone took place in early December 2018 and spent nearly an hour on two mathematical question chains.

Technical phone

  1. Start with the density of x1+x2 for uniform variables on [0,1], then extend it to the sum of n variables. Independence was not explicitly recorded.
  2. For an n×n matrix A other than I or −I, find the possible eigenvalues if A²=I, then if A^n=I.

H62: Uniform variables, matrix concepts and ML tools

Report date: 2018-11-22 · Role: Quantitative Researcher; seniority not stated

One phone covered the resume, probability and ML-related discussion. The candidate explicitly said there was no second phone.

Technical phone

  1. Find the distribution of X+Y when both variables are uniform on [0,1]. Independence was not stated.
  2. Define eigenvalues and eigenvectors and explain their use in ML.
  3. Which ML libraries have you used? Have you used TensorBoard?

H63: A different uniform-sum threshold and automatic outlier handling

Report date: 2018-11-07 · Role: Junior Quantitative Researcher

The interviewer was ten minutes late, and the call disconnected once. The interview itself lasted a little more than 30 minutes, including about ten on resume and ML work. The candidate thought they had failed.

Technical phone

  1. The interviewer went deeply into the candidate's resume and ML work.
  2. For N independent U(0,1) variables, find P(sum < 1/2), starting with small N and using induction. The candidate initially answered the familiar sum < 1 question before correcting the threshold.
  3. How would you deal with outliers in ML? After suggestions involving manual weighting and tuning, the interviewer asked for an automatic approach.

H64: Junior QR questions from volatility models to optimization

Report date: 2018-10-01 · Role: Junior Quantitative Researcher

The phone lasted 47 minutes. The candidate described an ML-heavy job description and said they had not heard further news at the time of the account.

Technical phone

  1. Which volatility models do you know? The candidate chose Heston, which led to its advantages and drawbacks.
  2. State regression assumptions and explain BLUE.
  3. What assumptions are used in financial or portfolio optimization?
  4. Prove a martingale result. The process and result were not retained.
  5. A gambler's-ruin problem followed, but the bankrolls, probabilities, boundaries and requested quantity were not recalled.

H65: A JQR-ML phone across probability, algebra and SVMs

Report date: 2018-04-05 · Role: Junior Quantitative Researcher, ML

The first technical phone lasted about 30 minutes and moved quickly through probability, statistics and linear algebra.

Technical phone

  1. For independent X and Y, find an upper bound on the standard deviation of X+Y. The fixed quantities or other constraints were not retained.
  2. What properties does a square matrix have if A²=A?
  3. Explain SVM and what to do when the data are not linearly separable.
  4. Explain Newton's method.
  5. Find the expected number of coin flips until two consecutive tails.
  6. How would you optimize an objective subject to constraints?
  7. Given a real sequence, how would you determine whether it has a limit?
  8. Explain the difference between independence and being uncorrelated.
  9. The candidate vaguely remembered one or two further questions.

H66: A later reply confirms conceptual phone interviews

Report date: 2018-03-26 · Role: Quantitative Researcher; seniority not stated

By an early-April update, the candidate had completed two phones, including a conceptual interview without live coding.

Second phone

  1. The candidate later confirmed two phone interviews. One involved conceptual questions without live coding; the other was described only by reference to a similar account, without its own question details.

H67: A research deep dive followed by regression and PCA

Report date: 2018-03-12 · Role: Quantitative Research Intern

The first phone began with an unusually detailed project discussion before turning to regression and PCA.

Technical phone

  1. A resume project was examined in considerable technical detail.
  2. State the assumptions behind linear regression and explain why they are needed.
  3. Explain and prove PCA's connection to eigenvalues and eigenvectors. Why does the formulation account for variance in X rather than variance in Y?

Keep practicing

In this series

Sources: Anywhere useful interview info can be found.

0 total reactions. Your reactions: none.
0 comments

Comments

0

No comments yet — be the first to share what you know.