RepostedJP Morgan Chase logo

JP Morgan Chase · OA → Phone screen → Onsite

In process

JPMorgan Quant Stats Superdays: 3 Experiences (Part 3 of 3)

Quant (general) · InternshipReported Jan 28, 2021
Shared byjanebeingjane

JP Morgan Chase quant intern Superdays from 2018 to January 2021: three to four 45-minute rounds on statistics, finance math, ML and light coding.

At a glance

All three reports in this part are older than three years, so they sit in History below. For 2021–2026 reports, see Parts 1 and 2.

Coverage: 3 JP Morgan Chase quant intern candidates (reports 17–19): two Quant Associate PhD interns (December 2020 and January 2021) and a Quantitative Research / ML master's intern (January 2018). Phone and OA steps stay inside each timeline.

In this series

  • Part 1: 2023–2026 reports, 1–9.
  • Part 2: 2021–2022 reports, 10–16.
  • Part 3 (current): 2018–2021 Superdays, 17–19.

Related JP Morgan Chase roundups: Tech intern · Data & AI intern

Patterns across the reports

  1. Four rounds, repeated topics. The two PhD candidates both had four interviews and saw topics repeat across interviewers: pi estimation and Black–Scholes came back in H17; behavioral and statistics questions repeated in H18.
  2. Your projects first. H19 said projects took more than half of the time; H17 opened by defending the candidate's research; H18 listed resume talk as one of the regular topics.

Common questions

How many rounds is the JP Morgan Chase quant intern Superday?

Four 45-minute interviews for the two PhD candidates (H17, H18); three 45-minute rounds in the 2018 report (H19).

Does the JP Morgan Chase quant Superday include coding?

Yes, but light: binary search and a two-table Pandas task (H17), short handwritten code (H18), and Pascal's Triangle and Median of Two Sorted Arrays on a whiteboard (H19).

What finance questions come up in the JP Morgan Chase quant intern interview?

Black–Scholes assumptions and delta, stochastic volatility, mean-reverting rates and the volatility smile (H17); Ito's lemma, derivative pricing and American options (H18).

History (before Nov 2023)

H17: Four 45-minute rounds on stationarity, Pandas and regression

Interview date: January 21, 2021 (report January 28, 2021) · Role: Quant Associate intern, PhD candidate · Result: Verbal offer January 25; team match January 27

Before the Superday

  1. Applied September 7 and got the OA the next day. The candidate did not finish one of the two coding questions and used the recorded explanation to say what went wrong.
  2. In October a separate data project arrived from an asset and liability management team. They sent an ARIMA-based forecast late and were not sure it was part of the same hiring process.
  3. A 45-minute phone interview on December 3: research, a three-dice DP puzzle (rules not fully recalled), Black–Scholes assumptions and delta, American versus European options, and estimating pi with Monte Carlo. After a hint, the candidate improvised an inscribed-polygon argument; the interviewer accepted the idea, though it is not a Monte Carlo method.
  4. The January 21 Superday had four 45-minute interviews with 15-minute breaks, and about 13–14 candidates that day.

Round 1: research and mathematical finance

  1. The interviewer challenged whether the candidate's mathematical-finance research was useful, and the candidate won him over.
  2. A delta calculation was stopped halfway because it was tedious. The interviewer mentioned a CDF/PDF link for Greeks that was new to the candidate.
  3. A difficult experience, stochastic-volatility models, stochastic-rate models and the implied-volatility smile. For rates the candidate could only recall mean reversion. The round ran over time.

Round 2: stationarity, a binary array and a two-table data task

  1. What must you check before fitting an AR or MA model? The candidate first missed it, then discussed stationarity and a unit-root test. Asked for a test name, they said Yule–Walker, which the interviewer rejected. Practice: Check Stationarity Before Time-Series Modeling.
  2. First One in a Sorted Binary Array: zeros followed by ones; find the first one faster than O(n). The candidate used binary search. The interviewer said she picked an easy task after hearing they were weak at data structures.
  3. Pandas: an employee table and a department table, both with employee_id; return the highest-paid employee in each department.
  4. A partly recalled puzzle with 1,000 coins and one unfair coin; Black–Scholes PDE derivation, assumptions and delta again.
  5. The candidate mentioned an ML course, so the interviewer asked what a random forest is and whether SVM suits many features with few samples, or few features with many samples. The candidate felt the ML answers went badly.

Round 3: estimation, outliers and feature selection

  1. Estimate pi again, with sampled points; this time the candidate knew the method.
  2. Explain maximum likelihood to a non-technical person.
  3. Regression outliers: the candidate proposed standardizing residuals and dropping points far from the mean. The interviewer questioned the precision and the normality assumption, then asked what to do when the outliers are not known in advance. Choosing important features followed.

Round 4: teamwork and too little complete data

  1. How the candidate completed an important task, and their role in a group project.
  2. Linear-regression assumptions and data checks (the author was unsure whether these were in round 2 instead).
  3. Missing data: delete rows or fill values. Then: what if deletion leaves too few rows to fit the regression?
  4. The verbal offer came January 25 and team matching January 27. The candidate later clarified the offer was from QR and model validation, not the data-project team.

H18: Four mathematical-finance interviews with repeated topics

Report date: December 15, 2020 · Role: Quant Associate intern, PhD candidate · Result: Not stated

  1. Four Superday interviews, each like the earlier phone interview.
  2. Each interviewer picked from resume talk, behavioral questions, statistics, mathematical finance, ML algorithms and short handwritten code.
  3. Statistics was around the level of a classic quant puzzle book. Stochastic calculus reached Ito's lemma, with derivative pricing and an American-option discussion in the last round, when the author felt worn out.
  4. Behavioral and statistics questions repeated across interviewers. The author chose not to share the individual prompts.

H19: QR/ML projects, Pascal's triangle and the median of two arrays

Report date: January 30, 2018 · Role: Quantitative Research / ML intern, master's candidate · Result: Not stated

  1. Online application, a phone interview a few weeks later, then an onsite invite about two weeks after that; under a month in total.
  2. The Superday had three 45-minute rounds, each starting with resume projects, then math and statistics or coding. Projects took more than half of both the phone and onsite time.
  3. Coding on a whiteboard: Pascal's Triangle and Median of Two Sorted Arrays.
  4. The third interviewer spent so long on projects that only basic math remained.
  5. The interviewers were friendly; the candidate could not tell how it went.

Keep practicing

In this series: Part 1 · Part 2 · Part 3 (current)

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.