HRT Python Track Interviews: 64 Experiences (Part 1 of 2)
- Company
- Hudson River Trading
- Reports
- 30 interview reports
- Level
- Unknown
- Stages
- OA → Phone screen → Onsite
HRT Python-track interviews from 2021 to 2026 test how Python works underneath, from dicts to generators, and often include debugging code over SSH.
At a glance
- Process: an OA or phone screen, then onsite rounds. The 2026 new grad phone screen was one timed coding task (20); the 2022–2023 new grad first rounds were knowledge and code reading instead (24, 25, 26).
- Asked recently: Python internals in 19 of the 30 Python-track reports, algorithm coding in 15, and debugging or explaining existing Python in 8. Generators and
yieldcame up in 8 reports across the series,withand context managers in 6, and how a dict works in 5. - Difficulty: in the 2026 maze task, reading every rule mattered more than speed; candidate 60 lost on one misread rule. Several Python candidates also got OS, memory and Linux questions (9 of 30).
- Practise first: Tilting Maze Ball, Debug Python scripts that read CSV files, Design a concurrent framework like Python's asyncio and the prep guide below.
Coverage: Hudson River Trading software engineering interviews · 30 Python-track or Python-tested reports (Software Engineer, Python Developer, Python Engineer, Operations Developer (Python), Data Software Engineer; new grad to senior, plus several with no level given) · 15 intern reports · 19 reports from SWE candidates whose language was C++ or not stated · US, UK, Canada and Singapore · August 2021 to October 2026
New grad on the Python track? Entries 20, 23, 24, 25 and 26 are new grad Python rounds, entries 28 and 29 are new grad Python OAs, and these, entry 27 and one more 2026 Python-track report that isn't in this series (nine new grad reports in all) are collected in one article: 2021–2026 HRT Software Engineer New Grad (Python Track + Python-Tested Roles) OA / Interviews.
In this series
- Part 1 (current): the 30 Python-track reports (Python-knowledge rounds, debugging over SSH, the full London Python loop, new grad Python rounds from 2021 to 2026, and two new grad Python OAs) plus the question tables for the whole series.
- Part 2: 15 intern reports, 14 onsite or final-round reports and 5 phone screens from C++ or unstated-language candidates, plus patterns across all 64.
What came up
Across the 30 Python-track reports, Python language knowledge is the most common thread. Nineteen candidates were asked about Python itself, beyond just writing code in it. Algorithm coding came up in fifteen. Debugging or explaining existing Python code came up in eight.
| Topic (Python-track reports) | Reports | Last seen |
|---|---|---|
| Python language knowledge and internals | 19 of 30 | Aug 2026 |
| Algorithm coding (LeetCode-style or puzzle) | 15 of 30 | Oct 2026 |
| OS, memory, networking, Linux or Bash | 9 of 30 | Jul 2026 |
| Debugging or explaining existing Python code | 8 of 30 | Aug 2026 |
| Background, past work, culture fit | 6 of 30 | Oct 2026 |
| System design | 2 of 30 | Feb 2026 |
| Take-home project with unit tests | 1 of 30 | Jan 2023 |
| Trade-data processing and trade data structures | 1 of 30 | May 2022 |
| SQL | 1 of 30 | May 2022 |
Inside "Python language knowledge", the dict is the repeat favorite: five separate candidates were asked how a Python dict works under the hood, and two more were shown a buggy dict subclass. Generators and context managers come up just as often. A candidate can show up in more than one row. For a spoken answer to each of these Python questions, see the companion prep guide: HRT Python-Track Interview Questions, Answered.
The reports from C++ or unstated-language candidates are in Part 2, including a 2026 new grad onsite. They show what the onsite day looks like more than what Python is asked.
Questions we found
These tables cover all 64 reports across both parts of this series. Each date is the interview date the source gave, or the post date where marked.
Python knowledge (spoken questions)
| Question | Reports | Last seen |
|---|---|---|
Generators, iterators and what yield is for |
8: 18, 21, 22, 24, 25, 26, 30, 44 | Jul 2026 |
How the with keyword and context managers work |
6: 06, 22, 24, 25, 26, 30 | Oct 2025 |
| How a Python dict works under the hood (hash collisions) | 5: 04, 09, 19, 21, 26 | Jul 2026 |
| How CPython manages memory, and how garbage collection works | 3: 04, 21, 30 | Jul 2026 |
| Explain decorators, including which ones can make code faster | 3: 21, 22, 27 | Jul 2026 |
| How the GIL works, and how a Python thread differs from an OS thread | 2: 18, 26 | Jun 2022 |
| Ten-plus short Python-fundamentals questions inside a timed OA | 1: 02 | Aug 2026 |
What * and ** mean |
1: 21 | Jul 2026 |
| How to make a class hashable, and whether a list can be a dict key | 1: 21 | Jul 2026 |
| How the interpreter and compilation handle a given piece of code | 1: 07 | Nov 2024 |
| Hash tables: collisions, cost of each operation, when and how they resize | 1: 18 | Dec 2021 |
Debugging and code reading
| Question | Reports | Last seen |
|---|---|---|
Spot the bug in a subclass of Python's dict (__setitem__) |
2: 12, 25 | Feb 2024 |
| Fix a program so its output matches the requirement | 1: 02 | Aug 2026 |
| Extend one API in a small existing codebase (intern) | 1: 31 | Aug 2026 |
| Read code with decorators, generators, context managers and threads: say what it prints, then fix or finish it | 1: 22 | Oct 2025 |
| Explain what a Python snippet does and how to improve it | 1: 08 | Oct 2024 |
Quick code reading: what a snippet does, what's wrong, how to optimize it (__setitem__ came up) |
1: 24 | Oct 2023 |
| SSH into a machine, find the error in a log file, fix the code or the data, then verify | 1: 24 | Oct 2023 |
Read short functions that use yield or with to read a file, explain them and improve them |
1: 25 | Oct 2022 |
| Check a short stock-price function for correctness and complexity, then improve it | 1: 25 | Oct 2022 |
| Debug Python scripts that read CSV files, with bugs in the data, the code or both | 1: 16 | May 2022 |
| A Python script got slower with no code change: why, and what would you do | 1: 16 | May 2022 |
| Fix a short Python script by changing only a few lines (OA) | 1: 16 | May 2022 |
Algorithms / Data Structures
| Question | Reports | Last seen |
|---|---|---|
| Union-find, with path compression | 4: 42, 43, 44, 49 | Jul 2025 |
| Build a deque (with index access, or circular) and a vector; pick the growth factor; is growth by +1000 still amortized O(1)? | 4: 34, 45, 59, 64 | Nov 2022 |
| A ball in a grid maze, moved by timed tilt events | 3: 20, 60, 61 | Oct 2026 |
| Contact-tracing graph problem: who may be infected after people meet over several days (intern) | 3: 32, 38, 39 | Nov 2024 |
| itoa: turn an integer into a string (also atoi, and watch 0 and the most negative int) | 3: 58, 59, 63 | Nov 2023 |
| Speed up 3D ray tracing: which surface or triangle does a ray hit first? | 3: 41, 42, 44 | Nov 2022 |
| Prices plus queries of (start index, budget): the most items you can buy from that index on | 2: 36, 56 | Sep 2026 |
| Build a hash map: collisions, a full table, then a balanced tree in each bucket | 2: 41, 42 | Nov 2022 |
| Find a local minimum (or peak) in an array, then in a 2D grid | 2: 34, 45 | Apr 2022 |
| Read any number of bytes through a fixed 4096-byte read call, across many calls | 2: 34, 58 | Apr 2022 |
| Remove adjacent AB or CD pairs until none are left (new grad Python OA) | 2: 28, 29 | Nov 2021 |
| In a tree whose nodes are labelled a or b, the longest path with no two equal neighbours (new grad Python OA) | 2: 28, 29 | Nov 2021 |
| The largest sum over diamond-shaped regions of a grid | 1: 61 | Oct 2026 |
| The street position lit by the most streetlights, smallest on ties | 1: 61 | Oct 2026 |
| A C++ reader for one file split into chunks, with starter code and a low-level API | 1: 57 | Aug 2026 |
| Two numbers that sum to a target; the K smallest elements; quickselect | 1: 62 | Jul 2026 |
| Does a root-to-leaf path sum to a target? Then return every such path (intern) | 1: 35 | Jul 2026 |
| A tree with a weight on each node: the largest minus the smallest weight for each node | 1: 36 | Sep 2025 |
| Fancy numbers; simulate a board game where placed pieces flip the opponent's pieces; merge two keyed trees and print the pre-order (intern OA) | 1: 40 | Sep 2024 |
| Find the polydivisible pandigital number | 1: 12 | Feb 2024 |
| Implement the 2048 game's move, using rotation so one direction covers all four | 1: 42 | Oct 2022 |
| Fewest changes to make every value in a sequence unique | 1: 15 | Jul 2022 (report date) |
| Does one string contain a fragment of another, under several constraints | 1: 15 | Jul 2022 (report date) |
| Pick at most 2 of N projects under ordering rules to get the most value | 1: 15 | Jul 2022 (report date) |
| Can a binary string be split into size-k blocks, each with a set number of 1s | 1: 17 | Feb 2022 |
| Repeatedly remove equal adjacent characters until none are left | 1: 18 | Dec 2021 |
| Build an n-ary tree from parent and value lists, then find the best k-length path sum | 1: 18 | Dec 2021 |
System Design
| Question | Reports | Last seen |
|---|---|---|
| Route network packets between one hub and several node servers | 1: 50 | Apr 2025 |
| A classic system design round (new grad onsite) | 1: 59 | Oct 2021 |
Low-Level Design (LLD/OOD)
| Question | Reports | Last seen |
|---|---|---|
| Schedule user-space threads or tasks without preemption, or build a coroutine step by step (intern) | 3: 37, 38, 39 | Mar 2025 |
| A tiered item store with expiry, levels and weights: init, store and retrieve (new grad onsite) | 1: 56 | Sep 2026 |
| How would you implement a garbage collector | 1: 47 | Nov 2025 (report date) |
| Design a concurrent framework like Python's asyncio (intern) | 1: 32 | Nov 2024 |
| A rate limiter in C++, with profiling and scaling follow-ups (intern) | 1: 34 | Oct 2023 |
| Hit counter that also returns each hit's payload, when payloads don't fit in memory | 1: 52 | Jul 2023 (report date) |
| A data structure for trade operations, built for speed | 1: 16 | May 2022 |
| A 2-hour onsite problem: fill in a class's functions, then optimize (lazy updates and similar) (new grad onsite) | 1: 59 | Oct 2021 |
SQL and data processing
| Question | Reports | Last seen |
|---|---|---|
| Process a set of trades and pull out the required information | 1: 16 | May 2022 |
| A SQL query with a JOIN, schema and sample data given | 1: 16 | May 2022 |
Take-home
| Question | Reports | Last seen |
|---|---|---|
| Build a small system in Python, with unit tests, in about a week | 1: 13 | Jan 2023 |
CS fundamentals
| Question | Reports | Last seen |
|---|---|---|
| Running many processes on one computer; ways processes communicate; how a shell pipe works | 6: 34, 42, 43, 44, 54, 59 | Oct 2022 |
| Threads vs processes, and why a thread switch is cheaper | 5: 21, 22, 26, 34, 45 | Jul 2026 |
| Stack vs heap, and what lives in each | 5: 18, 27, 30, 59, 62 | Jul 2026 |
C and C++ basics: which way the stack grows, inline, new vs malloc and writing your own malloc, what happens when malloc fails, segmentation faults, changing a const char*, little-endian checks |
4: 58, 59, 62, 64 | Jul 2026 |
| Ordered vs unordered maps, and how each is built | 4: 41, 43, 44, 58 | Nov 2022 |
| A trading system shows rare latency spikes or high latency variance: possible causes (memory, TLB, cache, page faults) | 3: 37, 38, 39 | Mar 2025 |
| Virtual vs physical memory, and how an address is translated | 3: 18, 41, 58 | Nov 2022 |
Linux: check memory, show the current directory, list files without ls, swap, symlinks vs hard links, delete a file named -rf, recovery after rm -rf |
1: 21 | Jul 2026 |
Is std::vector::push_back O(1)? |
1: 48 | Nov 2025 (report date) |
| What the OS does when you run a program from the command line | 1: 42 | Oct 2022 |
Reading a 100 GB file: mmap vs read; debugging a slow network |
1: 58 | Nov 2021 |
| UDP vs TCP | 1: 30 | Aug 2021 |
Behavioral (BQ)
| Question | Reports | Last seen |
|---|---|---|
| Why do you want to work here | 2: 53, 58 | May 2023 |
| Describe a recent interesting bug you fixed, or your hardest bug | 2: 18, 26 | Jun 2022 |
| Tell me about your past experiences | 1: 10 | Oct 2024 |
| Your favorite data structure, and where it helped you (intern) | 1: 34 | Oct 2023 |
| Culture-fit chat with the team head about the role and recent projects | 1: 16 | May 2022 |
| A time your own change broke production, and how you fixed it | 1: 58 | Nov 2021 |
| What matters most to you when choosing an offer | 1: 58 | Nov 2021 |
64 detailed candidate timelines, loops, and screens
Entries 01–30 are here; 31–64 continue in Part 2.
Python-track and Python-tested software roles
01: 30-minute coding exercise, C++ or Python
Report date: 2026-10-20 · Role: Software Engineer · Level: Junior · Result: No offer
Interview
- A talk about the candidate's background.
- A 30-minute coding exercise, explaining the approach in detail along the way. The candidate could pick C++ or Python. They called the problem interesting and challenging, but didn't say what it was.
02: OA packed with Python trivia
Report date: 2026-08-24 · Role: Senior Software Engineer · Level: Senior · Result: Did not pass the OA
Online assessment
- More than ten short questions on Python fundamentals. Some needed trivia-level knowledge.
- A coding part: fix a program so its output matches what the requirement asks for. Not hard on its own.
- The time was tight, and switching between the two kinds of questions ate into it.
03: CodeSignal, then a Zoom coding round with Python knowledge
Report date: 2026-02-04 · Role: Software Engineer · Location: New York · Result: No offer
CodeSignal (taken within a week of applying)
- Four questions. The candidate got three fully and didn't optimize the fourth.
Zoom coding interview
- Two or three LeetCode medium/hard questions.
- Some system design questions, plus questions on Python knowledge and data structures.
04: a CPython-internals round in the final
Report date: 2025-08-08 (interview "a while back") · Role: Not stated · Result: No offer after the final round
Final round
- One of the final-round interviews was on Python internals. It went into how CPython works, for example how a dict handles hash collisions and how memory is managed.
- The candidate spent a few weeks reading CPython source to get ready, and passed this interview. They didn't get through the final round overall.
05: two hours of Python syntax, live coding and debugging
Report date: 2025-04-21 · Role: Sr. Software Engineer · Result: No offer
Interview 1 (1 hour)
- Technical questions, live coding and debugging.
Interview 2 (1 hour)
- Much the same format. The candidate described the questions as Python syntax or straightforward LeetCode-style.
06: Python fundamentals deep dive, then a design-flavored coding round
Report date: 2024-11-13 · Role: Software Engineer · Level: Junior · Location: Singapore · Result: Did not pass the second interview
OA
- The standard CodeSignal test.
Phone screen
- A deep dive on Python fundamentals. One question: how does the
withkeyword work?
Second interview
- One coding question with a design focus. The candidate didn't describe it.
07: questions on how the interpreter works
Report date: 2024-11-01 · Role: Software Engineer · Level: Junior · Location: Singapore · Result: No offer
OA
- CodeSignal.
First round
- Deep Python questions. Some went into how compilation and the interpreter work, for example being shown some code and asked how the interpreter handles it.
08: Python Engineer round on syntax, Bash and debugging
Report date: 2024-10-14 · Role: Software Engineer (applied as Python Engineer) · Location: Singapore · Result: No offer
OA
- Not described.
Technical round
- Questions on Python syntax and Bash commands.
- Debugging-style questions, such as: here's some code, tell me what it does and how you'd improve it.
- The candidate enjoyed how the debugging questions were put together.
09: dict internals on the phone, then a five-part virtual onsite
Report date: 2024-10-09 · Role: Systems Software Engineer · Location: New York · Result: No offer
Coding round 1
- Four medium-level LeetCode problems.
Phone interview
- Resume discussion, plus basic Python and Linux questions. One was to explain how Python implements dictionaries.
Coding round 2
- Harder problems than round 1.
Virtual onsite
- Five conversations covering technical questions, system design and behavioral questions.
10: Python SWE in Singapore, full-day onsite and an offer
Report date: 2024-10-01 · Role: Software Engineer, Python · Location: Singapore · Result: Offer, accepted
CodeSignal
Two phone rounds (1 hour each)
Full-day onsite
- Scheduling took about two extra weeks because of availability.
- Interviewers explained what each round would cover before it started. The candidate felt they cared more about the thinking than about landing the exact answer.
- One question shared: tell me about your past experiences.
Results came back quickly after each round.
11: a Python plus JavaScript pre-screen
Report date: 2024-05-01 · Role: Software Engineer · Location: Chicago · Result: No offer
Pre-screen coding challenge (3 hours for both)
- A hiring manager reached out first.
- One Python problem and one JavaScript problem. Both leaned on language syntax more than algorithms. The JavaScript one was to design a game.
12: debugging over SSH, then a polydivisible pandigital number
Report date: 2024-02-01 · Role: Software Engineer · Location: US · Result: No offer
Interview 1: code reading over SSH
- The candidate SSHed into a VM and shared a terminal through tmux.
- They were shown Python snippets and asked to explain how each worked and spot possible bugs. One was a subclass of Python's
dict: find the bug that would cause an error. - One interviewer ran it, with two more people watching. The candidate found this round more technically involved than the next one.
Interview 2: CoderPad
- Traditional puzzle-style problems, nothing Python-specific.
- One was to find the polydivisible pandigital number. Polydivisible means each prefix of the number is divisible by its length. Pandigital means it uses each digit once.
- The candidate leaned on
itertoolsbuilt-ins. The interviewer seemed to expect the answer in another form, and the two didn't quite connect.
The recruiter then went quiet for two weeks.
13: a week-long Python take-home
Report date: 2023-01-01 · Role: Software Engineer (applied as Python Engineer) · Level: Junior · Location: Singapore · Result: Did not pass round 1
Take-home (about a week)
- Build a system in Python, with unit tests. The candidate didn't say what the system was.
14: three rounds of Python Q&A in one hour
Report date: 2022-10-12 · Role: Data Software Engineer · Location: UK · Result: Did not pass
Interview (1 hour)
- Three question-and-answer segments on basic Python.
- One LeetCode-medium BFS/DFS problem.
15: a Python Developer coding screen with three problems
Report date: 2022-07-25 (post date) · Role: Python Developer · Result: Not stated
Coding screen (OA)
- Find the fewest changes needed to make every value in a sequence unique. Solved.
- Decide whether one string contains a fragment of another, under several constraints the candidate didn't list. About 85% of the tests passed because of an off-by-one bug.
- A company has N projects with values V[N] and M ordering requirements between them, and can finish at most two projects. Pick the pair worth the most. The candidate ran out of time on this one.
16: the full Operations Developer (Python) loop in London
Report date: 2022-05-07 · Role: Operations Developer (Python) · Location: London · Experience: about 3.5 years at a large tech company · Result: Offer, negotiated, accepted; later joined
OA (1 hour, Codility)
- A SQL query with a JOIN, with the schema and sample data given.
- A short Python script with a bug, to fix by changing only a few lines.
- An easy Python algorithm problem.
Phone 1: debugging (1 hour, SSH into a prepared VM)
- Five scenarios. Each had a Python script and a CSV file. The bug could be in the data, in the script or in both: a bad file format, a syntax error and so on.
- The candidate finished all five in about 30 minutes.
Phone 2: data structures and algorithms (1 hour)
- Two questions. The first was around LeetCode medium and took most of the time.
- For the second, the candidate explained an approach but didn't get working code.
Virtual onsite (3 technical rounds + 1 culture fit)
- Debugging and troubleshooting (1 hour). Debug a Python script and its data file while screen-sharing and running it locally. Then an open question: a Python script suddenly got slower with no code change. What could cause it, how would you fix it, and which fix would you pick when?
- Trade data processing (1 hour). Process trade data and pull out the required information. More operations work than algorithms or design.
- Trade data structure (1 hour). Build a data structure that supports operations on trades, with the emphasis on speed and efficiency.
- Culture fit (45 minutes) with the global head of the team, about the role and recent projects.
17: a bit-manipulation-heavy OA for Python SWE
Report date: 2022-02-01 · Role: Software Engineering – Python · Location: US · Result: Did not pass the OA
OA (HackerRank)
- Three questions, rated LeetCode medium to hard, mostly bit manipulation.
- One example: given a binary string, decide whether it can be split into blocks of size k, each holding a specific number of 1s.
18: Python Developer OA and a phone screen full of internals
Report date: 2021-12-01 · Role: Python Developer · Location: Canada · Result: No offer
OA (Codility)
- Collapse a string by removing equal adjacent characters, again and again, until no pair is left. The candidate's examples:
ABBCDDCshrinks toA, andAABBAAshrinks to nothing. - Build an n-ary tree from a list of parents and values plus a number k, then find the largest sum of a path of length k.
Phone screen (45 minutes, Zoom)
In this order:
- Describe a recent interesting bug you fixed.
- What is
yieldfor? - How does a hash table work? What happens when two keys hash to the same spot? What do lookup, insert and delete cost? When does it grow or shrink, and what does that cost?
- How does memory work? How does the OS give memory to each process? Virtual vs physical memory, and what happens when virtual memory goes past physical?
- Stack vs heap, and what's stored in each.
- How does the GIL work, and what are its upsides and downsides?
19: dict internals and an array question
Report date: 2021-05-12 · Role: Software Engineer · Location: US · Result: No offer
Interview
- How a Python dictionary works under the hood.
- A basic algorithm question about deleting from an array.
- The candidate had seen some of the questions on review sites beforehand.
More Python-track reports
20: new grad Python phone screen, the same maze task as the C++ track
Report date: 2026-10-13 (post date) · Role: Software Engineer, New Grad (Python track) · Result: Passed the phone screen
Phone screen
- The candidate passed and was preparing for the second round.
- Moderation removed the details from their first post. In a reply, they pointed to a 2026 C++ new grad report (entry 61) and said they got the same task: a ball in a grid maze, moved by timed tilt events.
21: a Python and Linux knowledge round with no coding
Report date: 2026-07-15 (post date) · Role: Not stated (replies guessed an operations-leaning role) · Experience: Experienced · Result: Rejected about a week later
Phone screen (the recruiter called it Python and Linux basics)
- Python: explain decorators; how a dict is built and how it resolves collisions; what a generator is; what
*and**mean; how to make a class hashable; whether a list can be a dict key; how garbage collection works. - Linux: how to check memory on a host; the command that shows the current directory; how to list files without
ls; what swap is for; threads vs processes; symlinks vs hard links; how to delete a file named-rf; whether anything can bring files back after an accidentalrm -rf. - Partway through, the interviewer asked whether the candidate was using AI.
- In a reply, the author said HRT's phone stage was one knowledge round plus one coding round. Another candidate replied that they had a similar round plus an SSH task fixing code in vim, and were told the role leaned toward DevOps.
22: senior Python developer, code reading in CodeSignal
Report date: 2025-10-20 (post date) · Role: Senior Python Developer · Experience: MFE, working in quant · Result: Did not pass
First phone round (labelled CoderPad & Zoom; the code was in CodeSignal)
- Purely technical. No resume or behavioral questions.
- The interviewer showed full or partial code, asked what it does or prints, then asked the candidate to debug it or fill in the missing parts.
- Topics: decorators, generators, context managers, and threads vs processes. The candidate's advice is to know how to build and use each, the related functions, common examples and common mistakes.
- A second part was meant to SSH into HRT's server, but the candidate ran out of time before reaching it. Another candidate replied that their SSH part was simple Linux work: moving through several files.
23: new grad Python, what the recruiter said round 2 covers
Report date: 2023-10-15 (post date) · Role: Python Developer, New Grad · Result: Passed round 1; round 2 not reported
- The OA came first, then the phone rounds. The candidate applied directly and had interviewed for an HRT internship before.
- Round 1 matched another forum report (most likely entry 24). The candidate heard back 2–3 business days later.
- The recruiter email said round 2 covers operating systems, algorithms, data structures and Python internals.
24: new grad Python phone screen: code reading, then debugging over SSH
Report date: 2023-10-11 (post date) · Role: Python Developer, New Grad · Result: The candidate expected a rejection
Phone screen (1 hour, after passing the OA)
- Quick code reading. The interviewer shows a snippet; say what it does, what's wrong with it and how to optimize it. The candidate named generators, context managers and
__setitem__as concepts to review. - SSH into one of HRT's machines, find the error in a log file, change the code, then check the fix. If the data is wrong, fix the data. If the code is wrong, fix the code. The candidate compared it to an SRE support rotation.
- Medium difficulty. The candidate missed a few basic questions and spent too long on the debugging part, with many hints from the interviewer.
Earlier Python-track reports (2021–2022)
25: new grad Python, first round: read, fix and improve small functions
Report date: 2022-10-18 (post date) · Role: Software Engineer, New Grad (Python) · Result: Not stated; the candidate felt they answered poorly
First round after the OA (video)
- A short function using
yield: read it, explain it, then improve it. - A short function that reads a file with
with: the same, with basic knowledge questions along the way. - A function that computes stock prices: analyze its time complexity and correctness, point out problems, and say how to improve it.
- A class that inherits from
dict: find what's wrong. You need to know how__setitem__behaves.
In a reply, the candidate said "improve" meant two things: correctness (corner cases) and speed (can it be made faster).
26: new grad, a 45-minute round on how Python works
Report date: 2022-06-21 (post date) · Role: Software Engineer, New Grad (Core Developer listing; the recruiter described a Python-focused round) · Result: Not stated
First phone round (45 minutes)
- Your most interesting project and the hardest bug in it.
- What
yieldis, and iterators vs generators. - What
withdoes. - How a Python dictionary is implemented.
- Threads vs processes, and what each is good for.
- How a Python thread differs from an operating-system thread.
The questions took about 20 minutes. The candidate felt they got about 60% right.
27: new grad, a Python-sounding role with low-level questions
Report date: 2022-05-04 (post date) · Role: Market Structure Optimization Engineer (a software role), New Grad · Result: Not stated
Phone round
- What decorators can do, and which decorators can make code run faster (for example, caching).
- Heap vs stack.
- Ways to cut latency when talking to an exchange.
- From the candidate's ML project: GPU acceleration, and whether they'd design hardware to cut latency.
- Their biggest Python project, and how they know their code is correct.
The posting suggested a few Python packages would do, but the questions went low-level.
28: new grad Python OA: two string and tree problems
Report date: 2021-11-16 (post date) · Role: Python Developer, New Grad · Result: Not stated
OA (120 minutes, 2 problems)
- Remove adjacent "AB" or "CD" pairs until none are left. A stack works.
- In a tree whose nodes are each labelled "a" or "b", find the longest path where no two neighbouring nodes have the same letter.
29: new grad Python OA: the same two problems
Report date: 2021-10-28 (post date) · Role: Python Software Engineer, New Grad · Result: Not stated
OA (120 minutes, 2 problems)
- Remove adjacent "AB" or "CD" pairs.
- The longest path in an a/b-labelled tree with no two equal letters in a row.
The candidate rated it hard.
30: Python Core Developer, a phone screen on Python and networking
Report date: 2021-08-16 (post date) · Role: Python Core Developer (a role HRT described as fairly new at the time) · Experience: Experienced · Result: Rejected the next day
Phone screen (about 45 minutes)
- Python basics: what a generator is, context managers, garbage collection.
- CS knowledge: UDP vs TCP; heap vs stack in the operating system.
Common questions
What Python questions does HRT ask?
Mostly internals: how a dict handles collisions (04, 09, 19, 21, 26), generators and yield, context managers, decorators, memory management and the GIL. Candidates 12 and 25 were shown a buggy dict subclass to fix.
What is the HRT debugging round over SSH?
You connect to a prepared machine and read, run and fix Python there (12, 16, 24; 22 had an SSH part too), sometimes with bugs in the data as well as the code.
What is the HRT new grad phone screen in 2026?
One timed coding task: the tilting maze ball, given to a Python-track new grad (20), a C++ new grad (61) and an experienced candidate (60).
Keep practicing
- Practice from these reports: polydivisible pandigital number (12); fewest changes needed to make every value in a sequence unique (15); Pick the pair worth the most (15); a Python script and a CSV file (16); split into blocks of size k, each holding a specific number of 1s (17); Collapse a string by removing equal adjacent characters (18); find the largest sum of a path of length k (18); a ball in a grid maze, moved by timed tilt events (20).
- HRT banks: the coding bank and behavioral bank (sign-in needed), and every HRT question on the Hudson River Trading interview page.
- More reports: the Experience feed.
In this series: Part 1 (current) · Part 2
Prep guide: HRT Python-Track Interview Questions, Answered
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.