Nutanix Intern Debugging Round and Interviews: 25 Experiences

Shared by janebeingjane ·
Company
Nutanix
Reports
25 interview reports
Level
Internship
Stages
OA → Phone screen → Onsite

Nutanix SWE intern campus drives, mostly in India, ran a coding test, a written round on finding planted bugs, then technical and HR interviews.

At a glance

  • Process: the two reports from 2025 show a three-question online test, a 40-minute written debugging round, a technical round on DSA and OS, and a manager round (01); and a 40–45 minute technical round followed by HR (02).
  • Asked recently: candidate 01 had a digit-change divisibility question and "Gas station" in the test, a thread-safe BST full of planted bugs, a next/child linked list leading to running medians, an LRU cache on paper, connected components, and OS questions. Candidate 02 had heap and dynamic programming questions.
  • Difficulty: candidate 01 found 4 or 5 of the 7–8 planted bugs and needed hints on the running median; 02 felt slow at coding and credited their projects. Both got offers.
  • Practise first: Repair a Multi-Threaded Binary Search Tree, Running Median of a Multilevel Linked List in Level Order, LRU Cache and Count Connected Components in an Undirected Graph.

Coverage: 25 Nutanix software engineering intern reports from Aug 2016 to Aug 2025: MTS interns (including 6-month internships), SDE and software interns, mostly from campus drives in India (IIT BHU, IIT Kharagpur, IIT Kanpur, IIT Roorkee, NIT and unnamed colleges), plus a few U.S. summer interns. Only 01 and 02 are from the last three years; the other 23 (2016–2023) are under History, outside the current counts. U.S. reports have their own, larger roundup: Nutanix U.S. SWE intern interviews. Reports with no hiring level: Nutanix interviews, level not reported.

What came up

With only two current reports, the At a glance lines cover them. The longer pattern, from all 25 reports, is a separate debugging round on someone else's code, which 13 reports describe (see Patterns).

Questions we found

These tables cover the two current reports. The older questions are in the History tables.

Algorithms / Data Structures

Question Reports Last seen
LRU Cache, written on paper after a page-replacement discussion 1: 01 Aug 2025 (report date)
Digit Changes for a Multiple of 9: a 2 may become 4 and a 3 may become 9; can the number become divisible by 9? 1: 01 Aug 2025 (report date)
Gas Station Loop Start (named only as "Gas station") 1: 01 Aug 2025 (report date)
Running Median of a Multilevel Linked List in Level Order: level-order print of a next/child list, then running medians 1: 01 Aug 2025 (report date)
Count Connected Components in an Undirected Graph 1: 01 Aug 2025 (report date)

Debugging and code repair

Question Reports Last seen
Repair a Multi-Threaded Binary Search Tree: insert, delete, and search code with 7–8 planted bugs 1: 01 Aug 2025 (report date)

Behavioral (BQ)

Question Reports Last seen
Your favorite courses and topics, followed by questions on those topics 1: 01 Aug 2025 (report date)

CS fundamentals

Question Reports Last seen
Page replacement: what is wrong with FCFS, and why LRU does better 1: 01 Aug 2025 (report date)
Draw the life cycle of a process 1: 01 Aug 2025 (report date)
Virtual memory: how can a 5 MB video play on a device with 1 MB of RAM? 1: 01 Aug 2025 (report date)

2 candidate timelines from 2025

01: IIT BHU campus drive, offered a spot after the shortlist was already out

Report date: Aug 16, 2025 · Role: SDE intern (campus internship drive) · Location: IIT BHU, India · Result: Offer

Online assessment

  1. Three questions. Solving more than two of them was enough for the interview shortlist.
  2. A string of digits where some digits may be changed: any 2 can turn into a 4 and any 3 into a 9. Return whether some set of changes makes the number divisible by 9 (Digit Changes for a Multiple of 9).
  3. A LeetCode-medium style question he didn't describe.
  4. A question he named only as "Gas station"; our version is Gas Station Loop Start.

Debugging round (written, 40 minutes)

  1. A binary search tree with insert, delete, and search that also used threads. Candidates were told it held 7 or 8 bugs and had to mark and fix them. He found 4 or 5 and later thought this was why he missed the first shortlist. The task is set up as a repair project in Repair a Multi-Threaded Binary Search Tree.

Technical round (DSA and OS)

  1. A linked list where each node has a next pointer and a child pointer. Print it in level order; he used a queue.
  2. With a small change to the input, print the same structure in sorted order. His answer was an inorder traversal.
  3. Print the running median of the values from the first part, then produce the first part's output using the running median. He needed some hints on the last step, and the interviewer seemed satisfied. The level-order print and the running median are combined in Running Median of a Multilevel Linked List in Level Order.
  4. Caching and page replacement, including the weak spot of FCFS. He argued that LRU does better, so he was asked to write an LRU Cache on paper.
  5. A short talk about his projects, then his own questions.

Manager round

  1. His favorite courses and the topics he liked in them, followed by questions on those topics.
  2. Draw the life cycle of a process.
  3. Explain virtual memory. Then a practical version: how can a 5 MB video play on a device with only 1 MB of RAM?
  4. Write code to count the connected components of a graph.

The interviewer was a friendly alumnus. Nutanix first picked only two interns from this campus. In the second week the placement office called him: the company had decided to take one more person, and he got the offer.


02: IIT Kharagpur SDE and R&D intern, selected mostly on projects

Report date: Jul 20, 2025 · Role: SDE and R&D intern · Location: IIT Kharagpur, India · Result: Offer

Technical round (40–45 minutes)

  1. Data structures questions on heaps.
  2. Dynamic programming questions.

HR round

  1. The interviewer was already interested in his CV and asked him to walk through his web and game development projects.
  2. A conversation about the company and the role.

Only circuit branches with a CGPA of at least 8 could apply. He had done little competitive programming, felt slow when coding in the technical round, and believes his unusual projects carried him.


History (2016–2023)

These 23 reports are more than three years old. They show the loop in full detail, and many of their questions are on FastPrep. Each table counts these 23 reports only.

Algorithms / Data Structures: asked in more than one report

Question Reports Last seen
Minimum Time to Publish Comic Books with printing and binding machines 2: 06, 07 Aug 2021 (report date)
Maximum Value of the Last Robot when robots knock out their nearest neighbor 2: 06, 07 Aug 2021 (report date)
Preorder of Subtrees in an N-ary Tree for many query nodes 2: 19, 21 Sep 2018 (report date)
Office to Home Walk: one chosen road, then one random road, per move 2: 19, 21 Sep 2018 (report date)
Reach a Cell With Sum Moves from (1, 1) 2: 19, 21 Sep 2018 (report date)
Reign of Every Element: the widest range where each element is the maximum 2: 19, 21 Sep 2018 (report date)

Algorithms / Data Structures: asked once, 2019–2023

Question Reports Last seen
Distance Sums to Special Nodes in a Weighted Tree, with a follow-up for K special nodes 1: 06 Aug 2021 (report date)
Water Supply With Wells and Pipes (approach and pseudocode only) 1: 06 Aug 2021 (report date)
LRU Cache, designed inside a system design round 1: 08 Jan 2021
Volunteer Awards: Totals Divisible by K: four parts, from subsets to subarrays to range queries with updates 1: 08 Jan 2021
Flip Rows and Columns, Count Zeros in a binary matrix 1: 08 Jan 2021
Longest Increasing Path in a Matrix 1: 08 Jan 2021
Nutanix Sports Meet: how many groups reach the finish on a one-lane road 1: 10 Aug 2020 (report date)
Team Tournament: longest prefix of fights that still allows a two-team split 1: 10 Aug 2020 (report date)
Maximum Combined Distance Across Five Arrays 1: 11 Aug 2020 (report date)
Paint Fence Work Split (named only as "paint fence") 1: 16 Dec 2019
Grumpy Owner's Calm Stretch (written as "grumpy broker") 1: 16 Dec 2019
Print All File and Folder Paths 1: 16 Dec 2019

Algorithms / Data Structures: asked once, 2016–2018

Question Reports Last seen
A binary search problem with two query types (our closest practice version: Rebuild the Line From Taller-in-Front Counts) 1: 20 Aug 2018 (report date)
A tricky variant of Count Inversions (the twist wasn't reported) 1: 20 Aug 2018 (report date)
Ternary Expression to Binary Tree 1: 20 Aug 2018 (report date)
Ways to Decode 1: 20 Aug 2018 (report date)
Minimum Starting Points to Cross a Grid 1: 22 Sep 2017 (report date)
Count Undominated Points 1: 22 Sep 2017 (report date)
Recover a BST by Swapping Two Nodes Back (swap the nodes, not the values) 1: 22 Sep 2017 (report date)
Count Components With a Prime Total Value 1: 22 Sep 2017 (report date)
Packed Bit Array: Set and Reset Bits (a sieve squeezed down to one bit per number) 1: 22 Sep 2017 (report date)
Lockable Nodes in a Binary Search Tree 1: 22 Sep 2017 (report date)
Delete a Linked-List Node Given Only Its Pointer (secondhand: asked to other candidates) 1: 22 Sep 2017 (report date)
Mother Vertex in a Directed Graph (secondhand) 1: 22 Sep 2017 (report date)
Find All Repeated Numbers (secondhand) 1: 22 Sep 2017 (report date)
Sum String Check 1: 25 Aug 2016 (report date)
Shuffle a String by Swapping Tree Children 1: 25 Aug 2016 (report date)
Largest Minus Smallest Value of a Parenthesized Expression 1: 25 Aug 2016 (report date)

Debugging and code repair

Question Reports Last seen
Repair a Readers-Writer Lock 2: 06, 25 Aug 2021 (report date)
Repair Circular Linked List Insert and Split 2: 19, 21 Sep 2018 (report date)
Repair Multi-Threaded File Merging with a Waiting Queue: threads copy several input files into one output file 2: 19, 21 Sep 2018 (report date)
Repair Ordered Acknowledgments for Concurrent File Writes 1: 08 Jan 2021
Repair a Linked-List Merge Sort That Removes Duplicates 1: 20 Aug 2018 (report date)

System Design

Question Reports Last seen
Design a Pizza Chain Ordering and Delivery System (asked as a Domino's-style delivery system) 1: 06 Aug 2021 (report date)
Design a Pastebin Text Sharing Service 1: 08 Jan 2021
An Instagram-style photo platform (secondhand: another finalist's question; practice: Social News Feed With Live Engagement) 1: 19 Sep 2018 (report date)

Behavioral (BQ)

Question Reports Last seen
Tell me about yourself / introduce yourself 3: 04, 06, 21 Jul 2022
Interests and hobbies (one interviewer also asked for a favorite TV show) 3: 06, 19, 21 Aug 2021 (report date)
Where would you like to be located? 2: 19, 21 Sep 2018 (report date)
Why are you interested in competitive programming and software development? 1: 04 Jul 2022
Do you prefer working from home or in the office, and why? 1: 04 Jul 2022
Your path through college, and how you got into machine learning 1: 04 Jul 2022
Explain one of your projects so that someone without a technical background can follow it 1: 19 Sep 2018 (report date)
Why this company, and what are your future plans? 1: 21 Aug 2018 (report date)
How would you handle a teammate who isn't performing? 1: 21 Aug 2018 (report date)
Tell me about your favorite project 1: 22 Sep 2017 (report date)

03: IIT Kharagpur summer intern, debugging folded into the technical interview

Report date: Jul 30, 2023 · Role: SDE summer intern · Location: IIT Kharagpur, India · Result: Offer

Coding round

  1. Cleared; he did not name the problems.

Technical interview

  1. He describes Nutanix's debugging task as Python code in which you find the bugs. In his case it was merged into the technical interview.

Nutanix came on Day 1 of the first phase of the campus internship season.


04: IIT Kharagpur Software and Systems profile, two rounds and a call five minutes later

Interview date: Jul 31, 2022 · Role: Software Engineering Intern (summer 2023) · Location: IIT Kharagpur, India · Result: Offer

Round 1: technical (about 45 minutes, two interviewers)

  1. After introductions, a 10-minute debugging task in a Google Doc. The interviewers first explained what the snippet should do. The errors were hard to spot, and the candidate still finished inside the time.
  2. A medium-level competitive programming question. First the brute-force idea and its time complexity, then an optimized approach, then code for it, explaining each part and walking through small example inputs.
  3. OOP concept questions until time ran out.

Round 2: HR (about 25 minutes)

  1. Their time at the institute and how they got into machine learning through a course they took.
  2. Why are you interested in competitive programming and software development?
  3. Questions on leadership and on how they would act in certain situations, plus some open-ended, creative ones.
  4. Do you prefer working from home or in the office, and why?

The candidate was on the extended Day 1 shortlist and interviewed at noon. The offer call came about five minutes after the HR round.


05: NIT, 6-month MTS intern, problems that weren't standard

Offer date: Oct 28, 2021 · Role: MTS Intern (6 months) · Location: India (on campus) · Result: Offer

  1. Coding questions on a segment tree, dynamic programming, and n-ary trees. The candidate didn't remember the exact problems and said none of them looked like standard ones.
  2. A separate OS debugging round, with no details given.

The round order wasn't stated. A reply adds that for the batch graduating a year earlier (2022), Nutanix hired for intern plus full-time, with an extra system design round for final-year students; that wasn't part of this candidate's own loop.


06: Virtual campus drive, one offer out of thirteen

Report date: Aug 24, 2021 · Role: MTS intern · Location: Virtual, on campus · Result: Offer

Coding round (HackerRank, 90 minutes)

  1. Publish X comic books with Y printers and Z binders. Each printer has its own time per comic, every binder takes K minutes, and each machine handles one comic at a time. A comic may wait any amount of time between printing and binding. Find the earliest time all X are done. This candidate passed 20 of 21 tests. It is on FastPrep as Minimum Time to Publish Comic Books.
  2. Robots stand in a row, each with a value. A robot can remove its nearest neighbor on either side, and its value becomes its own value minus the removed one. Find the highest value the final remaining robot can end with. All tests passed (Maximum Value of the Last Robot).

Thirteen candidates made the cut.

Debugging round (Zoom, 45 minutes)

  1. Each candidate got their own document with readers-writers synchronization code. Every function's intended behavior was spelled out, and the job was to find the logic errors, much like Repair a Readers-Writer Lock.

Seven cleared it. Two of them then left for other offers, so five reached the technical round.

Technical interview (Zoom and a Google Doc, 60–75 minutes)

  1. Introductions, interests, hobbies, and a favorite TV show.
  2. A tree of N nodes with weighted edges and one special node X: return the sum of distances from X to every node. The candidate treated it as a graph and ran BFS.
  3. Follow-up: now there are K special nodes. For each node, add up its distances to all of them, and return the smallest total and the node that has it. With hints, the candidate seeded a BFS with all K nodes at once and tracked the smallest distance; the interviewer accepted it. Both parts are in Distance Sums to Special Nodes in a Weighted Tree.
  4. A new problem: every house needs water. Digging a well at house i has its own cost, and a pipe between two houses has an edge cost. Give the cheapest plan, as an approach, pseudocode, and complexity only. A greedy idea over sorted well costs was wrong; after a hint, the candidate built a minimum spanning tree and dropped edges that cost more than a well. Practice: Water Supply With Wells and Pipes.

Three moved on to system design.

System design (Zoom, 1 hour)

  1. Design a Domino's-style pizza delivery system, starting from a basic structure. The follow-ups fell into a few groups: requirements and data (functional and non-functional needs, the database choice, the table schema, and the structure behind the database); security (user privacy and the protocol); growth (scaling once the database fills up, faster responses, and which cache to use and how to build it); and sessions and load (expiring logins through stored timestamps, and a flood of requests). Our practice version is Design a Pizza Chain Ordering and Delivery System.

The candidate had spent only 3–4 hours on system design beforehand, and the interviewer gave hints when needed. This candidate was the only intern selected.


07: Same Aug 2021 online assessment, posted as screenshots

Interview date: Aug 2021 (online assessment) · Role: MTS intern (6 months) · Location: Virtual, on campus · Result: Not stated

Online assessment (HackerRank, 1.5 hours)

  1. The comic book printing and binding problem, here with several test cases per input and three worked examples.
  2. The robot elimination problem, with N up to 500,000 and values between −10^9 and 10^9.

The post lists a CGPA cutoff of 7.5. It gives the same two problems as entry 06; the two posts don't show whether they came from the same campus.


08: Off-campus MTS intern, a four-part follow-up and a Pastebin design

Interview dates: Jan 15–19, 2021 · Role: MTS intern · Location: Virtual, off campus · Result: Offer (5-month internship)

Coding round (75 minutes, Jan 15)

  1. An n × m binary matrix starts as all zeros. Queries either flip a whole row or column, or ask how many zeros a row or column holds; up to 10^5 queries. A brute-force attempt passed about half the tests, and the candidate stopped early because of another commitment. Practice: Flip Rows and Columns, Count Zeros.
  2. The longest strictly increasing path in a matrix, moving only to side neighbors. DFS with memoization passed everything.

Debugging round (45 minutes, Jan 16)

  1. Around 20 candidates worked at the same time, each in their own document. The code handled concurrent file writes that had to be acknowledged in order. Only logic faults counted, not syntax, and the problem statement described what each part should do. The candidate flagged 10 or 11 faults. Reading the code carefully, they felt, mattered more than OS theory. Nine people went through. The task is set up as Repair Ordered Acknowledgments for Concurrent File Writes.

First interview: DSA (HackerRank CodePair, 1 hour, Jan 19)

One setup with four questions. A row of houses each holds some money. Volunteers collect for three NGOs, and a volunteer whose total can be split evenly three ways earns an award. Many volunteers may visit the same house, but no volunteer visits a house twice.

  1. Volunteers may pick any houses: count the subsets whose sum is divisible by 3. The candidate coded a DP, then answered the follow-up for k NGOs.
  2. Volunteers may only take consecutive houses: count the subarrays whose sum is divisible by 3, using prefix-sum remainders.
  3. Answer up to 10^5 range queries (l, r) on the second version. The candidate used a segment tree, which the interviewer hadn't expected.
  4. Add updates that change one house's amount; the same segment tree handled them.

All four parts are in Volunteer Awards: Totals Divisible by K. The interviewer was pleased that they got through all four.

Second interview: system design (1 hour, Jan 19)

  1. Design Pastebin for a given list of requirements: API signatures, which kind of database with its trade-offs, its collections, and caching.
  2. Within the caching discussion, the design of an LRU cache.

The candidate had only skimmed a few blog posts on system design just before the interview. The interviewer guided them, and they finished on time.


09: U.S. Summer 2021 software intern phone screen

Interview date: Oct 2020 · Role: Software intern (Summer 2021) · Location: U.S. · Result: No feedback yet

Phone interview

  1. Coding at about LeetCode-medium level; no problem was named.
  2. Conceptual questions based on their background and experience.

10: Summer Internship 2021 campus test, posted as screenshots

Report date: Aug 26, 2020 · Role: Summer Intern 2021 · Location: On campus (college not named) · Result: Not stated

Online assessment

  1. "Nutanix Sports Meet": runners start at different points on a one-lane road toward a finish M miles away, each with a top speed. No one can pass the runner ahead, so a faster runner who catches up joins that group. Count the groups that reach the finish.
  2. "Team Tournament": N fighters and an ordered list of fights. Every fighter is placed on RED or BLUE, and no fight may be between teammates. Fights run in order until one more fight would make such a split impossible; return how many fights can happen.

The author passed only a few test cases on each. A reply says it was the on-campus internship test.


11: Intern online assessment with five arrays

Report date: Aug 18, 2020 · Role: Intern · Location: An IIT, per a reply · Result: Not stated

Online assessment

  1. Five arrays of the same length: find the largest value of the sum of |a[i] − a[j]| across all five arrays plus |i − j|. In replies that appear to be the author's, the absolute values were split into sign cases; a later reply generated all 32 sign patterns with a bitmask instead of writing them out by hand. Practice: Maximum Combined Distance Across Five Arrays.

12: Bangalore MTS intern, an interviewer who kept asking for more

Offer date: Spring 2020 · Role: MTS Intern (6–7 months) · Location: Bangalore, India (on campus) · Result: Offer

  1. Online round (HackerRank, 90 minutes): two medium-to-hard questions, not described.
  2. Technical interview 1, debugging: fix the logic errors in a C program for an OS task. It was real code, not pseudocode, and the candidate says solid OS fundamentals were enough.
  3. Technical interview 2: a DSA question that combined Dijkstra with DP.
  4. Technical interview 3: a question that could be solved with strongly connected components. After the optimal answer, the interviewer kept asking for something faster. The candidate came up with other approaches of the same complexity and was asked to code one; it felt like a stress test.

13: U.S. MTS intern, San Jose, Spring 2020

Offer date: Spring 2020 · Role: MTS Intern · Location: San Jose, CA · Result: Offer

  1. Coding screen: two questions, not described.
  2. Technical round: by phone.
  3. HR round.

The candidate was an M.S. student whose only prior work was one internship.


14: M.S. summer intern assessment that ended early

Attempt date: Feb 2020 · Role: Summer Intern (M.S.) · Location: Not stated · Result: Not stated

  1. Proctored online assessment: the proctoring setup (browser extensions, mic and camera) hung Firefox on a recent Mac. Switching to Chrome about ten minutes in set off a too-many-tabs warning, and the test was ended.

No questions were described. The candidate suggests using Chrome.


15: Software internship phone interviews, four coding questions

Interview date: Dec 2019 (replies on Dec 24, 2019 and Jan 27, 2020) · Role: Software intern · Location: Not stated · Result: Not stated

Phone interview(s)

  1. Two coding questions; the interviewer tested the code on a few cases.
  2. A few OS questions.
  3. In a later reply, the same person (most likely) remembered four coding questions in total: one on bit manipulation, one on arrays with a follow-up, and one on binary trees.

It isn't clear whether the four questions span both phone interviews.


16: U.S. Summer 2020 intern, coding challenge and a live file-path question

Interview date: Dec 2019 · Role: Summer 2020 intern · Location: California · Result: Not stated

Coding challenge (proctored, 90 minutes, video and audio on)

  1. A question the author called "paint fence", most likely the fence-splitting question from another 2019 Nutanix OA report: Paint Fence Work Split.
  2. A question the author called "grumpy broker", most likely the grumpy bookstore owner problem (Grumpy Owner's Calm Stretch).

Technical interview (45 minutes, recorded)

  1. About 15 minutes of résumé questions.
  2. Code live in C++ or Python and print every file and folder path, such as /, /combinations, /combinations/locker, /foo/a. The input format wasn't given. Practice: Print All File and Folder Paths.

Hiring manager interview

  1. Listed as the third step, with no details.

The author warns that the proctoring software flags even a little background noise.


17: Software internship, Era database team, résumé-only first call

Interview date: Dec 2019 · Role: Software intern (Nutanix Era database team) · Location: Not stated · Result: In process

First phone interview

  1. Résumé discussion and technical questions about it, with no data structures questions. The interviewers expected the candidate to ask about the team.

HR's instructions suggested the second call would be behavioral, but after the first one a second technical round was scheduled, later pushed to early February.


18: U.S. MTS intern, San Jose, January 2019 offer

Offer date: Jan 18, 2019 · Role: MTS Intern · Location: San Jose · Result: Offer

  1. Coding screen: two questions, not described.
  2. HR round.
  3. Two technical rounds: both by phone.

The candidate was a master's student with a year of prior MTS R&D experience.


19: 2018 campus drive for a 6-month internship plus a full-time role (first of two reports)

Report date: Sep 26, 2018 · Role: 6-month intern + full-time offer (software engineering) · Location: On campus (college not named) · Result: Offer (6-month intern + full-time, as part of the HR group)

Online test

  1. Many queries over an n-ary tree, each naming a node: print the preorder traversal of that node's subtree. The tests forced precomputing every answer in a single pass over the tree. Practice: Preorder of Subtrees in an N-ary Tree.
  2. A city graph with an office and a house. On each move, a woman first picks a road on purpose, and if she isn't home yet, she then takes a random road other than the one she just used. Can she ever get home with the best choices, and if so, in how few moves? The author passed 8 test cases (Office to Home Walk).

Only students above an 8.5 CGPA could take the test, and 16 passed it.

Debugging round (paper)

  1. A C program implementing a circular doubly linked list, including splitting one list in half; see Repair Circular Linked List Insert and Split.
  2. A Java program where several threads read from several files and write into one file, close to Repair Multi-Threaded File Merging with a Waiting Queue.

Six survived the paper round.

Technical round 1 (45 minutes)

  1. Go from (1, 1) to (n, m), with n and m up to 10^7, where each move goes from (x, y) to (x, x + y) or (x + y, y). Print the path if there is one. DP and BFS were too slow for these limits, so the author worked backward from the target. Practice: Reach a Cell With Sum Moves.
  2. After solving, an open conversation about projects, the company, and some technologies.

Each technical round had a single question, and only a correct answer kept a candidate in the process.

Technical round 2 (45 minutes)

  1. For every array element, find its "reign": the widest stretch of the array in which it is the maximum. The author used a stack and finished in about 20 minutes (Reign of Every Element).
  2. Another relaxed conversation after the solution.

HR round

  1. The HR interviewer had no technical background and asked the author to explain a project so that someone without that background could follow it.
  2. Questions on interests and location preference.

Five finalists were split: three went to HR and two to a system design round. Secondhand, from the author: one of those two was asked to build an Instagram-style platform (closest practice: Social News Feed With Live Engagement), and the other got a database question the author couldn't recall. The HR group received the 6-month internship plus a full-time offer; one person from the design group got the internship only.


20: Campus intern drive with a two-page written debugging task

Report date: Aug 17, 2018 · Role: Software intern · Location: On campus (college not named) · Result: Offer

Coding test (1 hour)

  1. A binary search problem with two kinds of queries, each needing a different binary search. An O(n²) solution passed 6 of 9 tests, and an O(n log n) one passed all. Our closest practice version: Rebuild the Line From Taller-in-Front Counts.
  2. A tricky variant of counting inversions; the author didn't describe the twist and ran short on time. Practice for the standard count: Count Inversions.

Fully solving one problem was enough to qualify. Eleven people went on.

Written debugging round

  1. Two pages of code that sort a linked list with merge sort and drop duplicates. Only mistakes in the main logic counted, not missing semicolons or pointer dereferences. The author caught 4 or 5 real ones. The practice project is Repair a Linked-List Merge Sort That Removes Duplicates.

Seven were left after it.

Technical interview

  1. Turn a ternary expression such as a?b?c:d:e into a tree (Ternary Expression to Binary Tree). It took about 25–30 minutes with code.
  2. A dynamic programming question: Ways to Decode. Solved quickly.

HR round

  1. No details, and the author was selected for the internship afterward.

21: Same 2018 campus drive, second report

Report date: Aug 10, 2018 · Role: 6-month intern + full-time (software engineering) · Location: On campus (college not named) · Result: Not stated

Online round (HackerRank, 1.5 hours)

  1. A tree with up to 10^5 nodes and as many queries; every query asks for the preorder traversal of one node's subtree. Practice: Preorder of Subtrees in an N-ary Tree.
  2. An undirected graph that has a start node and a target node. Each minute, take a road on a shortest path, stop at the target, and otherwise take a random road other than the one just used. Find the worst-case time to arrive, or −1. This report states the goal differently from entry 19. Practice: Office to Home Walk.

Fourteen got through.

Debugging round (75 minutes)

  1. Code that adds a new head node to a circular linked list, plus code that cuts one circular list into two of half the length. Practice: Repair Circular Linked List Insert and Split.
  2. OS code where several threads each read their own input file and all write to one output file, at most 400 characters at a time, with threads that can't get the file waiting in a queue. Practice: Repair Multi-Threaded File Merging with a Waiting Queue.

That left six.

Face-to-face round 1

  1. From (1, 1), reach (M, N) on an unbounded grid, with M and N up to 10^7, each move going to (X + Y, Y) or (X, X + Y). Print the path, or NO. The author's tip: go backward (practice: Reach a Cell With Sum Moves).
  2. Then questions about projects, or sometimes C or Java.

Face-to-face round 2

  1. For each number in an array, give the range where it is the largest element, in O(N). The author's tip: use a stack. Practice: Reign of Every Element.
  2. Again followed by project or language questions.

Everyone in a round got the same question at the same time. Five students solved both rounds optimally. Three went to HR, and two had an extra system design round with all three interviewers together.

HR round

  1. Questions on the résumé, an introduction, why Nutanix, hobbies, plans for the future, and where they'd like to work.
  2. How would you handle a team member who isn't performing?

The three HR candidates got a 6-month internship plus a full-time offer, and one of the two from the extra round got the internship alone.


22: Campus intern drive with parallel panels and a bit-level sieve

Report date: Sep 17, 2017 · Role: Software intern · Location: On campus (college not named) · Result: Not stated (2 of the final 6 were selected)

Coding round (1.5 hours)

  1. The minimum positive points needed to cross a grid. The author's forward DP turned out to be wrong and passed 3 tests.
  2. Count the points that no other point dominates, where a point dominates another if both of its coordinates are larger. All tests passed.

About 130 people took the test and 12 were interviewed.

Interview round 1 (three panels in parallel)

Each batch of three candidates got the same questions, and the questions changed for the next batch. Secondhand, from other batches: Delete a Linked-List Node Given Only Its Pointer, Mother Vertex in a Directed Graph, and Find All Repeated Numbers in an n-element array holding values 0 to n − 1. A few easier questions the author didn't remember were also asked.

The author's own questions:

  1. Tell me about your favorite project: the author described one on orthogonal Latin squares.
  2. A BST in which two nodes were swapped: find them and swap the nodes themselves back, not just their values (Recover a BST by Swapping Two Nodes Back). The author coded it.
  3. In an undirected graph where every vertex carries a value, count the connected components whose total value is prime. Practice: Count Components With a Prime Total Value.
  4. Implement a sieve. The interviewer kept pushing on memory: why an int array, can a bool array do, how many bits does a yes or no need? He then gave a struct with an int array pointer and a value z, asked how big the array must be for 10^6 bits (the answer was 10^6 / 32 ints), and asked for code to set bit z and then to clear it. That last part is Packed Bit Array: Set and Reset Bits.

Six of the twelve went on.

Interview round 2 (technical + HR)

  1. A BST whose nodes have left, right, value, and locked fields. Given a node, decide whether it can be locked: no ancestor and nothing in its subtree may be locked. The author first walked down to the node checking ancestors and then scanned the subtree, then sped it up with an extra field recording whether a subtree contains a locked node, and coded that. It is on FastPrep as Lockable Nodes in a Binary Search Tree.
  2. A few general HR questions.

23: IIT Kanpur, August 2017, later on the Calm team

Interview date: Aug 2017 · Role: Software intern (later placed in the Nutanix Calm team) · Location: IIT Kanpur, India; internship in Bangalore · Result: Offer

  1. Coding round (1 hour): two questions, not described.
  2. Debugging round: fix a C++ snippet printed on paper.
  3. Personal interview: questions about projects and some data structures and algorithms.

Teams were assigned at random, though a preference mentioned in the interviews might count a little.


24: IIT Roorkee, 2017 internship season, also placed in Calm

Interview date: 2017 internship season · Role: Software intern (later placed in the Nutanix Calm team) · Location: IIT Roorkee, India; internship in Bangalore · Result: Offer

  1. Online coding test (about 1 hour): two questions, not described.
  2. Debugging round: a paper with code, mostly C or C++, where every kind of error had to be found and fixed.
  3. Interview: a few data structures and algorithms questions; the rest followed the candidate's interests. There was no separate HR round.

Nutanix visits on the first or second day of the internship season.


25: 2016 campus intern drive with a timed debugging interview

Report date: Aug 31, 2016 · Role: Software intern · Location: On campus (college not named) · Result: Not stated

Round 1: coding (75 minutes)

  1. Decide whether a digit string longer than 3 is a "sum-string", meaning it splits into numbers where each one equals the sum of the two before it. "12358" qualifies (1 + 2 = 3, 2 + 3 = 5, 3 + 5 = 8), and so does the string you get by writing 1, 99, 100 and 199 in a row; "2368" does not. All tests passed.
  2. Decide whether s2 can be built from s1 by splitting s1 into a binary tree and swapping the children of some nodes (for example, "great" to "taerg"). Five tests passed. Practice: Shuffle a String by Swapping Tree Children.

Twelve qualified.

Round 2: paper (about 30 minutes)

  1. A digit string with only + and * operators: add parentheses to get the largest value and the smallest value, and return the difference. For 2*2+2 that is 8 − 6. The author wrote an interval DP on paper, and the interviewer tested it on small cases and liked the clean, commented code. Practice: Largest Minus Smallest Value of a Parenthesized Expression.

Six were shortlisted.

Round 3: debugging

  1. A modified readers-writers lock. The author found five problems: a deadlock, a write-after-write inconsistency, a read-after-write inconsistency, a wrong if/else branch, and a syntax error, then wrote the fixed code in about 10 minutes. The interviewer was timing the author with a stopwatch. A similar repair task: Repair a Readers-Writer Lock.

Patterns across all 25 reports

  1. A debugging task, usually as its own round after the coding test. 13 reports describe one, all from campus or virtual drives, most of them in India (01, 03–06, 08, 12, 19–21, 23–25). Its form varied: code on paper (01, 19, 20, 23, 24), a separate shared document per candidate during a video call (06, 08), a snippet inside the technical interview (03, 04), a timed face-to-face session (25), and rounds whose format wasn't described (05, 12, 21). In 8 of them the code was an OS program, most often with threads, locks, readers and writers, or several file writers (01, 05, 06, 08, 12, 19, 21, 25). Three candidates were told up front what the code was meant to do (04, 06, 08). They disagreed on what helped most: one felt that careful reading beat OS theory (08), while another said solid OS fundamentals were enough (12).

  2. The debugging round cut deep. Five reports give numbers before and after it: 13 to 7 (06), about 20 to 9 (08), 16 to 6 (19), 11 to 7 (20), and 14 to 6 (21).

  3. One problem, then harder versions of it. Several interviewers kept building on the same setup: a next/child linked list that turned into running medians (01), tree distance sums that grew to K special nodes (06), the four-part volunteer question that ended with range queries and updates (08), a sieve that had to shrink to one bit per number and a lock check that had to get faster (22), and an interviewer who kept asking for more after the optimal answer (12).

  4. System design showed up even for interns. The two 2021 virtual-drive candidates (06 and 08) both had a full system design round, pizza delivery and Pastebin, and both said they had barely prepared and leaned on the interviewer's hints. In the 2018 drive (19, 21), only two of the five finalists got a design round.

  5. Everyone in a batch got the same question. In the 2018 face-to-face rounds, all students were asked the same question at the same time (21), and in 2017 three parallel panels asked one batch of three the same set before switching (22).

  6. The U.S. reports, and those with no stated location, had a different loop. Seven reports (09, 13–18) mention only coding screens, proctored tests, and phone, video, HR, or hiring-manager interviews. Three of them give the full sequence of a coding screen followed by interviews (13, 16, 18); the rest cover one or two phone interviews (09, 15, 17) or a test that ended early (14). Four are explicitly in the U.S. (09, 13, 16, 18), and none of the seven describes a debugging round. Two candidates mention strict proctoring: noise warnings (16) and a test that ended over open browser tabs (14).

Common questions

What is the Nutanix debugging round?

You get code someone else wrote, often an OS program with threads and locks, and must find and fix planted bugs. Candidate 01 (2025) had a thread-safe BST with 7–8 bugs on paper in 40 minutes; older reports describe paper, shared-document and timed face-to-face versions (Patterns, point 1).

How many students get through the Nutanix campus drive?

Few. Five older reports give numbers before and after the debugging round, from 13 to 7 (06) down to 16 to 6 (19). Candidate 01's campus first picked only two interns and later added one more.

Do Nutanix interns get a system design round?

Sometimes. Both 2021 virtual-drive candidates had one, pizza delivery (06) and Pastebin (08), and in the 2018 drive only two of five finalists did (19, 21).

Keep practicing

Sources: Anywhere useful interview info can be found.

0 total reactions. Your reactions: none.

Comments

0

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