2026 Zscaler India Campus SDE Prep: 37 Past OA / Interview Experiences (Part 1 :)

Question bank: 27 Coding/DSA + 16 C/C++/Systems · 14 links to 13 FastPrep problems · Last seen: 2019–Nov 2025.

Coverage: India campus, university, and college hiring. The completed reports are from 2019–2025. They cover SDE and Software Engineering interns, Intern + FTE/PPO, Associate Software Engineer, C/C++, Java/API, DevTest, UI, QA, SRE/DevOps, and Cloud Operations.

A practical prep order: timed coding mocks, spoken DSA practice, networking, OS/Linux, language internals, then two projects and six honest behavioral stories.

What candidates have reported in 2025 so far

1. One Software Development Intern candidate said their college skipped the OA. Zscaler used a resume shortlist and moved candidates to interviews.
2. A reply from another college claimed four medium OA questions, then one easy and one medium live coding problem plus OOP, OS, CN, and DBMS. The same commenter had earlier mixed up Zscaler and ZS Associates, so this detail is low confidence.
3. Several Skybound candidates said they had finished Round 1 and were waiting for Round 2 or a TPO update. They did not share the questions.
4. Another campus discussion mentioned two or three rounds and a roughly 20-minute discussion after them. Again, no question list was posted.
5. A low-context comment mentioned Floyd–Warshall, subset enumeration, and bit manipulation. The role and campus cycle could not be confirmed, so it is not treated as a verified Zscaler campus question.

Online assessments (OA)

OA 01: Two DSA questions and four aptitude questions

Report date: November 2025 · Role: SDE Intern · Process: On-campus · Platform: HackerRank

1. The test had two DSA questions and four aptitude MCQs.
2. The candidate solved one DSA problem and partly solved the other.
3. The candidate could not remember the two prompts.

OA 02: Java/C++ section plus a required Python section

Interview date: August 2025 · Role: Software Engineer Intern · Process: University · Duration: 2 hours

1. The OA had four questions.
2. Section 1 had two DSA questions in Java or C++.
3. Section 2 had two Python questions.
4. The candidate did not clear the OA. The report did not include the prompts.

A separate late-2025 candidate described the same broad split: a DSA/backend section in a preferred language and a required Python section. That second report was not clearly on-campus, so it is not part of the count.

OA 03: Four questions split across Python and C/C++

Report date: April 2025 · Role: SDE Summer Intern · Process: On-campus · Duration: 90 minutes

1. Two questions had to be solved in Python. The other two had to be solved in C or C++.
2. The candidate remembered a grid shortest-path problem. The report says they used DFS, but it does not give the full grid rules.
3. Another task used a greedy array approach to maximize profit.
4. A third task used a priority queue to build an efficient client-server network.
5. The report does not preserve the fourth prompt.
6. Around 350–400 students took the test. Sixteen moved to interviews.

OA 04: Java, interfaces, SQL, and fundamentals in one test

Interview date: May 2024 · Role: Java Developer Intern · Process: On-campus · Duration: 120 minutes

1. The OA had nine questions.
2. One medium greedy problem had to be solved in Java.
3. One Java task required the candidate to implement supplied interfaces.
4. Two SQL tasks used joins, sorting, grouping, and limits.
5. Five questions covered aptitude, OOP, and DBMS.
6. Twelve candidates moved to interviews.

The same drive also hired C and Python developers. This format was for the Java track.

OA 05: C/C++ tracks ranged from topic-only to hard competitive programming

Dates: 2023–2024 · Roles: Dev C and Software Developer C/C++

1. One 2024 Dev C test had three HackerRank coding questions in about two hours. The prompts were not saved.
2. Another Developer C report listed four OA areas only: HashMap implementation, Trie, Greedy with a Heap, and DP based on Longest Increasing Subsequence.
3. A 2023 Intern + FTE HackerRank OA had four DSA questions in two hours; the prompts were not saved.
4. A 2023 C/C++ test had three medium-to-hard questions in 90 minutes: BFS plus Binary Search, DP on Trees, and Segment Tree plus Binary Search.
5. That candidate solved the first two. For the tree DP, they used the centroid of the tree. They ran out of time on the third question.

OA 06: UI and DevTest tracks had web-specific sections

Dates: 2023 · Roles: UI Developer, UI Frontend Developer, UI DevTest

1. One UI OA lasted about two hours. It had HTML/CSS/JavaScript MCQs, three coding questions, and one DOM task. The candidate passed all but one DOM test case.
2. Another UI OA had 11 questions in 90 minutes: three coding questions, including one JavaScript task and two modified LeetCode Medium tasks; one DOM development task; and frontend MCQs.
3. A UI DevTest OA had one DSA problem restricted to Java or JavaScript, one JavaScript implementation task, and seven HTML/CSS/JavaScript MCQs.
4. The original reports did not publish the full coding prompts.

OA 07: Older drives often mixed coding with 20–35 MCQs

Dates: 2020–2023 · Roles: ASE, QA, Cloud Operations, DevTest

1. The MCQs covered CN, OS, Linux/Unix, DBMS, cloud services, code output, and puzzles.
2. Coding tasks included array and matrix work, scheduling, strings, ciphers, graph cycles, bit manipulation, and Python-only function work.
3. Some reports say candidates advanced after solving two or three coding questions. A 2023 ASE report also says some people solved all four but failed a plagiarism review.
4. These reports explain why core CS still appears often. They do not prove that a 2026 OA will use the same format.

Live interview question bank (compiled from all the 37 candidates reports)

FastPrep links point to the closest honest practice version.

Coding and DSA

1. Remove the Nth node from the end of a singly linked list. The 2025 candidate solved it in a live HackerRank CodePair session. The report does not say whether a dummy node or one-pass solution was required.
2. Maximum path sum in a binary tree. The same round asked for the largest sum over a path that may start and end at any nodes. No input limits were saved.
3. Minimize a sum with a heap. The interviewer said it was heap-based. The candidate only reached a partial solution, and the exact operation on the values was not saved.
4. Print a Longest Common Subsequence. The candidate could compute the LCS length but could not reconstruct the sequence correctly. Practice the length-only DP foundation.
5. Rotate a linked list. The report names the standard medium problem but does not save the rotation rules or constraints.
6. Array and sorting task. The candidate needed one hint and then finished. The prompt is missing.
7. Merge overlapping intervals. One candidate coded an O(n log n) solution and reported O(1) extra space. The code ran and the interviewer accepted it.
8. Three Sum to a target. The candidate explained a correct two-pointer approach. Time ran out before coding, and the interviewer accepted the explanation. Practice the standard 3Sum form.
9. Odd-even linked list by node position. Put nodes at even positions before nodes at odd positions. The candidate asked whether “even” meant the stored value or the position, and also checked empty-list and one-node cases.
10. Search in a rotated sorted array. The interviewer did not name the pattern. The candidate first had to notice that the array looked rotated.
11. Bit-position sum. For 4 (0100) and 3 (0011), the differing positions were 0, 1, and 2, so the requested sum was 3. The candidate used XOR in a brute-force attempt, but only two tests passed.
12. Count values in B that are no greater than each value in A. The interviewer wanted only the approach. The candidate took 10–15 minutes and reached sorting plus binary search.
13. Find the value with odd frequency. The report gives only the problem name.
14. Sum the divisors of the GCD of all array values. The report gives this task shape but no constraints or examples.
15. Run-length encoding in C. One report only says it had to be written strictly in C. A detailed 2020 version banned maps and extra arrays, gave spare space at the end of the same input buffer, and required aaaabbbca to become a4b3c1a1.
16. Build palindromes from two strings. Generate possible palindromes, combine them, and return the longest result; break a tie by lexicographic order. The candidate first misunderstood the task, so the interviewer could not ask the second planned DSA question.
17. Asteroid Collision-style stack problem. The report says it was similar to the standard problem but does not preserve the changed rules. The linked practice page uses the conventional collision rules.
18. Count vowel strings with adjacency rules. The candidate had to count length-n strings made from a, e, i, o, and u, with rules such as a only following e and e only following i. The full set of transitions was not saved, so the linked practice page clearly states its completed transition table.
19. C string pattern matching. The candidate first explained and coded an O(n²) brute-force solution. The interviewer then asked for an optimization, and the candidate explained Rabin–Karp.
20. Custom malloc and free in C. The candidate had not seen the problem. With one or two hints, they designed and coded what they described as an O(1) solution.
21. Detect and remove a linked-list cycle. This was followed by a recursive Fibonacci function and a time/space analysis.
22. Add two integers without using plus. The candidate had to use bit operations.
23. Validate an IPv4 string in C. The candidate had little C experience and finished with interviewer guidance. The source does not give the validation cases.
24. A long linked-list operation problem. The candidate only partly solved it before time ran out. The exact operations were not saved.
25. String compression where single letters also get a count. The output must write a1, not just a. The supplied buffer was twice the input length.
26. Queue and large-log nth-state problem. The author remembered a complex queue, very large logs, and a print function for the nth state, but explicitly said the exact task was no longer remembered.
27. String subsequence problem. The report names string subsequences but does not preserve the operation, objective, or requested output.

C/C++ and systems questions

1. Implement runtime polymorphism in C++. The candidate ran the code, and the interviewer accepted it.
2. Explain when a vtable is created, how virtual dispatch works, and why a virtual destructor matters. One candidate could not recall the virtual-destructor details.
3. Explain auto, extern, static, and register: scope, lifetime, and storage. One candidate also gave examples such as using static to keep a function inside one file.
4. Explain malloc, calloc, realloc, and free. One follow-up asked whether realloc extends the same block or moves the data. The candidate reasoned that an in-place extension could collide with another block.
5. Compute the sizes of two supplied structs, then explain padding and alignment in terms of how the processor reads memory.
6. Explain dangling, void, const, and volatile pointers. Another interview covered pointer size, pointer-to-const versus const pointer, valid pointer arithmetic, and why two pointers cannot simply be added.
7. Draw the byte layout of a hexadecimal value in little-endian and big-endian order. Then write a function that detects the machine’s endianness.
8. Given IP1, IP2, Mask1, and Mask2, decide whether the addresses are in the same network. The candidate handled classful addresses but not the classless case.
9. Explain unordered_map hashing, collision chains, and rehashing. A follow-up removed rehashing, used a small bucket count, and required O(log n) operations. The candidate reasoned toward a tree per bucket; the interviewer was pointing to ordered maps and red-black trees.
10. Prove the time complexity of binary search and the correctness of Floyd’s tortoise-and-hare cycle detector.
11. Implement a custom malloc/free in C. The candidate reached a working answer with hints.
12. Explain constructor and destructor order across base and derived classes.
13. Explain process memory: code, data, heap, and stack. Compare stack and heap allocation.
14. Explain who calls main and walk through preprocessing, compilation, assembly, linking, and execution.
15. Inspect whether a selected byte is nonzero. One interviewer asked for pointer arithmetic and byte-sized types. The candidate could not give a precise method. The linked exercise uses a portable numeric representation rather than raw pointer access.
16. Implement and modify small examples of inheritance, polymorphism, virtual functions, and function overloading. The report does not preserve the exact class scenarios.

Networks and operating systems

1. A request gets no response. How do you tell whether the machine is down or only the destination port failed? One candidate proposed reverse proxy and load-balancer tracking. The interviewer expected traceroute.
2. Explain TCP versus UDP with real use cases. Then explain TCP flags, retransmission, and the three-way handshake.
3. Explain a TLS handshake, certificate use, public/private-key encryption, and the move from TCP setup to a secure session.
4. Walk through DNS, DHCP, and ARP. Explain why DNS commonly uses UDP and when TCP is used.
5. Explain IPv4 versus IPv6, subnet masks, subnet IDs, CIDR, gateways, routing versus switching, MAC addresses, and NICs.
6. Explain personal VPNs versus enterprise VPNs, including tunneling, encryption, and authentication.
7. Explain what happens after entering a URL or clicking a link.
8. Explain what happens when an Ethernet cable is connected and how dynamic IP assignment works.
9. Compare a process and a thread. Cover shared resources, scheduling, context switching, and whether an ordinary function call creates a new thread.
10. Explain race conditions, critical sections, semaphores, mutexes, deadlock, paging, virtual memory, page replacement, and thrashing.
11. In a single-core machine with 1 GB of memory, three processes of 526 MB each are running. Explain how the OS can handle it.
12. Explain parent, child, and zombie processes.
13. For SRE/DevOps, walk through Linux boot, memory management, common commands, and a troubleshooting case. One selected candidate had no live DSA in either face-to-face round.

Java, SQL, frontend, projects, and behavior

1. Compare interfaces and abstract classes. Explain inheritance, overload versus override, the diamond problem, and functional interfaces.
2. Explain Java threads, synchronized blocks, semaphores, exception handling, collections, static/final, and Java memory.
3. Compare eager, lazy, synchronized, and double-checked-locking Singleton implementations.
4. SQL tasks covered joins, Group By, sorting, limits, table creation, keys, constraints, and returning the employee name and ID with the maximum salary.
5. Frontend candidates were asked how JavaScript executes, how Promise and async work, how to write a Promise, and follow-up questions on that code.
6. Other frontend topics included React error boundaries, polyfills, debouncing, null versus undefined, local versus session storage, and editing supplied React code in CodePair.
7. UI DevTest interviews asked for UNIX and Git command syntax, clone versus fork, pull request versus push, merge-conflict handling, HTTP status codes, Postman CRUD, and the APIs used in projects.
8. Project follow-ups asked for the candidate’s own contribution, architecture, tech choices, database choice, test plan, failures, and what they would change.
9. Behavioral prompts covered leadership, teamwork, a weak skill area, an unhelpful or unrecognizing manager, college struggles, why Zscaler, why security, and what the candidate would change at their college.

36 detailed candidate timelines, loops, and screens

01: 2025 SDE intern — two DSA questions, then two technical rounds

Report date: November 27, 2025 · Role: SDE Intern · Process: On-campus · Result: Rejected

Online assessment

1. HackerRank had two DSA questions and four aptitude MCQs.
2. The candidate solved one DSA problem and partly solved the other.
3. They could not remember the prompts.

Technical interview 1 — Google Meet and HackerRank CodePair

1. Remove the Nth node from the end of a singly linked list.
2. Find the maximum path sum in a binary tree.
3. Discuss Java multithreading, synchronization, and parallelism at pseudocode level.
4. Explain personal contributions to resume projects and answer basic OS questions.

Technical interview 2 — Google Meet

1. Solve a “minimize sum” problem. The interviewer said it was heap-based.
2. Solve a bit-manipulation problem involving non-primitive data types. The candidate could not recall the exact task.
3. The candidate reached only partial solutions and felt the interviewer was not satisfied.

The candidate received no follow-up, even after trying email and LinkedIn.

02: 2025 Software Engineer intern — LCS reconstruction and linked lists

Report date: April 29, 2025 · Role: Software Engineer Intern · Process: On-campus · Result: No offer

1. The candidate cleared the OA. The report says the full interview loop had three rounds.
2. In the first interview, they were asked to print a Longest Common Subsequence. They could compute the length but could not reconstruct the sequence correctly.
3. They compared a circular linked list with a circular queue.
4. They worked on linked-list rotation.
5. The last problem used arrays and sorting. The interviewer gave one hint, and the candidate finished it.
6. The candidate did not move to round two.

Part 2 continues with timelines 03–09.