Jobra
🐞Step 1 of Jobra:AI Debugging Rounds
DSA+AIPreparing you for both.
JOBRA · AI DEBUG ROUND (PREVIEW)sum.cpp
// buggy snippet
int total(vector<int>& a, int n) {
    int sum = 0;
    for (int i = 0; i <= n; i++) { // 🐞        sum += a[i];
    }
    return sum;
}
A real codebase. A real bug. You + AI.

Jobra drops you into a broken repo. Find it, fix it, ship it. AI helps you debug like on the job.

Prompt:Correct the loop condition
Tap 'Let Jobra fix it' below to simulate AI debugging...

Atlassian Software Engineer Interview Questions & Rounds

36 min read

Atlassian software engineer interview guide: the OA or Karat screen, coding, code design, system design and values rounds, from real candidate experiences.

In this article
On this page

[Draft] How to Crack Atlassian: The Software Engineer Interview, From Real Experiences

Last updated November 20, 2026 · 27 min read · For interns, new grads (P30), P40 and P50 candidates

Built by the Codinzhub team. This guide is built from Atlassian software engineer interview experiences that candidates shared, including experiences and OA questions on Codinzhub. The process changes by year, level and team, and it changed noticeably in 2026, so treat this as a map, not a promise.

The 30-second version. The Atlassian Software Engineer interview starts with a 90-minute HackerRank OA of 3 or 4 coding questions for interns and campus graduates, or a Karat screen (about 5 rapid-fire design questions plus 1 or 2 coding problems) for experienced hires. The loop then has up to two 60-minute coding rounds (Data Structures and Code Design, or an AI-enabled coding round in 2026), a 60-minute system design round from P40 up, and values and management interviews of about 45 to 60 minutes. You code in your own IDE, and follow-ups extend the problem. A hiring committee and a team match call follow. The habit that matters most: get working code running, test it, then extend it.

What this guide is, and what it is not

This guide is built from 51 candidate reports of Atlassian's software engineer interviews: 6 experiences and 4 online-assessment question pages shared on Codinzhub, plus 41 public posts, checked against Atlassian's own interview guides. 29 of the 51 are from 2025 and 2026; older campus reports, back to 2020, serve mainly as question examples. They cover interns, graduates (P30), P40 and P50, mostly in India.

It is:

  • A map of the process, stage by stage, as candidates described it
  • The question patterns that keep coming back, and the follow-ups to expect
  • A six-week plan and practice problems matched to those patterns

It is not:

  • A list of leaked questions or a promise of what you will be asked
  • An official description of Atlassian's process
  • An offer rate. Outcomes in candidate posts are self-reported and are not a statistical sample.

Important things to know before you start

These are the Atlassian-specific facts that change how you prepare.

You code in your own IDE, and you are expected to test. Atlassian's backend interview guide asks for your IDE and testing framework, a shared screen, and code you actually run; it recommends test-driven development. Five candidate reports confirm their own IDE. Practise in a real editor with tests.

Search is for syntax; AI only where the round allows it. In one 2025 report Google was allowed but AI was not, and another 2025 candidate who searched for the algorithm itself was not recommended. The AI-enabled round, with a coding agent in a HackerRank repository, appears only in 2026 reports.

Experienced loops are staged and can down-level you. Atlassian schedules the loop in three stages with a recruiter check after each, and three 2025 reports describe moving down a level mid-process.

Round-by-round breakdown

Candidates described five stages: application, shortlisting, an online assessment (campus and early careers) or a Karat or engineer screen (experienced hires), the interview loop, and a hiring committee with team matching.

Stage 1: Application, where candidates applied

Campus hiring in India dominates the intern and graduate reports. Experienced-hire reports rarely say how the candidate applied.

ChannelWhat candidates said
Campus drivesThe most common route (10 of 15 reports that name a channel): resume through the placement cell, a shortlist, then an OA link a few days later.
Careers siteOne intern applied online, got an OA link within days, then waited about a month for the interview invite.
ReferralOne SDE III candidate in Seattle was referred and went straight to interviews.
Recruiter outreachOne P40 candidate with 3 years of experience was approached directly by a recruiter.
Events and off-campusOne 2021 intern came through Atlassian's Women in Tech event; one 2026 graduate was hired off campus with a CGPA of 7.

What helped. Candidates credited projects on the resume, a referral that skipped the screen, and being ready to demo a deployed project.

Stage 2: Shortlisting, how candidates moved forward

Campus candidates were shortlisted on eligibility and resume before the OA; careers-site applicants described a questionnaire.

  1. Eligibility and resume shortlist (campus). Two drives limited the OA to CSE, ECE and EE students with a CGPA above 7; a 2026 IIT Delhi intern drive allowed CSE, MnC and EE. One older drive shortlisted about 100 students on CV and CGPA. One intern believed OA solvers who switched tabs, had no projects or had a CGPA below 7.5 were cut, but was unsure about the CGPA point.
  2. Questionnaire (careers site). One 2025 applicant got numbered yes/no questions on salary, relevant experience, work authorisation and sponsorship, then heard nothing.
  3. Recruiter calls between rounds. One 2026 intern got an unscheduled 20-minute recruiter call right after the AI-enabled coding round.

What candidates advised. Keep real projects on your resume and do not switch tabs during the proctored OA. One intern called these two points "very important".

Stage 3: Online assessment or phone screen

Interns and campus graduates took a timed online assessment (OA), a coding test on HackerRank; Atlassian's early-careers guide says clearing it is mandatory. Experienced hires (P40 and above) took a live screen instead, usually run by Karat, a third-party interviewing company.

Online assessment

How it works. The 9 reports that give a length say 90 minutes. Ten of 16 reports that give a count had 3 coding questions, five had 4, and one 2024 intern report had 2. HackerRank is the only platform named (10 of 20 reports). One 2022 drive ran it with camera and microphone on. The bar was high: one graduate drive interviewed only students who solved everything (20 of 317), and intern drives needed at least 2.5 of 3 (2024) or 3.5 of 4 (2022).

At a glance

  • Length: 90 minutes
  • Questions: 3 coding questions in most reports, 4 in some
  • What was asked: an easy string, maths or binary-search question, then DP, graphs (one with disjoint sets) and trees; the 2026 IIT Delhi intern OA had tree DP, DP on an array and an IP request-blocking system
  • Most common: No single repeated question
QuestionWhat it testsLevel
Request-blocking system for IP addressesSet and ordered-multiset bookkeepingIntern
Add and renew items with expiry times, then count unexpired itemsHash map simulation over timeIntern
Repeatedly delete every occurrence of string b from string aStack-based string reductionCampus
Smallest window that contains all characters of a stringSliding window with countsIntern
Split a numeric string into the fewest prime partsDP over prefixes with a sieveNew grad

OA questions shared on Codinzhub.

See all Atlassian OA questions in the question bank.

Karat or technical screen

How it works. Seven of 9 screen reports describe Karat. Two give the same timing: about 5 minutes of introductions, 20 minutes of rapid-fire design questions, then about 30 minutes of coding, with a second coding problem only if the first went quickly. The design questions were short scenarios, not full diagrams. Three reports mention a redo or retry. In 2026 the format moved: one India recruiter said Karat would now be 1 design question plus 2 coding questions, and two candidates were told their screen would be 45 minutes with an Atlassian engineer, 20 minutes of coding and 15 of design.

At a glance

  • Length: about 55 minutes on Karat (two reports); 45 minutes for the 2026 engineer-led screen (two reports)
  • Questions: 5 rapid-fire design questions plus 1 or 2 coding problems in most Karat reports
  • What was asked: design scenarios (find the issue, scale it, estimate capacity, model the data) and coding built on hash maps, BFS or DFS
  • Most common: No single repeated question
QuestionWhat it testsLevel
Given a slow or failing system, name the issue and fix itCaching, load balancing, fault toleranceP40
Estimate server size from daily active users and growthCapacity estimationP40
Model an order object with foreign keys and filter on themRelational data modellingP40
Pros and cons of keeping data on the device versus the serverTrade-off reasoningP40

Stage 4: The interview loop

Atlassian's backend interview guide (P30 to P50) lists two 60-minute coding interviews, a 60-minute system design interview, a 60-minute management interview and a 45-minute values interview, in three stages with a recruiter check after each. Experienced candidates described these parts, with the coding rounds named Data Structures and Code Design. Campus loops were shorter: one or two technical interviews, sometimes a design round, then one combined management and values (M&V) round. In 2026, interns and one senior backend loop had an AI-enabled coding round.

What a 60-minute coding round looks like

TimeWhat usually happensWhat to do
0–5 minIntroductions, sometimes a project questionHave one project ready to explain
5–15 minProblem read out; candidates asked about constraints, then gave a brute forceAgree on one example and say the brute force and its cost
15–40 minInterviewer asks you to optimise; you code it in your IDEName the data structure, then code in small pieces
40–55 minEdge cases, then follow-ups or "scale-ups" that extend the problemRun a test for each extension
55–60 minYour questionsAsk about the team or the work

One senior candidate wrote that you need to solve both problems with their extensions to get an offer, and two P40 candidates who did not finish a scale-up were unsure they had passed.

Coding and data structures interview

At a glance

  • Length: 45 to 60 minutes (60 in 9 of 14 reports that give a length); Atlassian's guide caps it at 60
  • Questions: 1 main problem with follow-ups in several reports (10 of 29); some campus rounds had 2 or 3 short problems
  • What was asked: hash maps, trees and graphs, intervals and DP; campus rounds added projects and CS fundamentals
  • Most common: minimum rooms or platforms for overlapping intervals, and a content popularity tracker (2 reports each)

You code in your own IDE with your screen shared (five reports and Atlassian's guide), or on a CodePair or HackerRank pad. In one 2025 P40 report, Google was allowed for syntax but AI was not. Five campus reports spent time on projects, and five reports asked CS fundamentals such as CPU scheduling, DBMS, networking, Java threading or HashMap internals.

QuestionFollow-upLevel
Minimum rooms or platforms so no intervals overlapAfter a heap solution, redo it with a sweep lineIntern
Content popularity tracker with O(1) increase, decrease and most popularState insert and delete cost of an ordered mapIntern
Merge k sorted lists, then with a multiset instead of a heapHow a multiset and red-black tree workSDE III
Last day you can still cross a flooding gridNot reportedSDE III
Lowest common ancestor in a tree you first build from a diagramNot reportedP40
Longest streak with constant daily change, then allow one editChange at most one elementP30
Fill '?' in a binary string so no palindrome of length 5+Not reportedP40
Implement an LRU cacheRequirements had to be clarified firstCampus

Code design interview

At a glance

  • Length: 60 minutes (Atlassian's guide)
  • Questions: 1 design-and-code problem with scale-ups (5 of 6 reports)
  • What was asked: small in-memory systems: a snake game, a file system, an agent rating system
  • Most common: Snake game (2 reports)

You design a few classes and write runnable code, then the interviewer adds requirements. One P40 candidate could not get runnable code for the scale-up; another ran out of time on an extensible design.

QuestionFollow-upLevel
Snake game on a grid: move, grow on food, dieAdd food, then teleportingP40, Senior
File system with directory sizes, then a top-N directories reportPrecompute the report or compute it on requestSenior
Agent rating system with running averages, then monthly averagesGroup ratings by monthNot reported

AI-enabled coding interview

At a glance

  • Length: 60 minutes (one 2026 report)
  • Questions: 1 task in an existing multi-file repository: fix a bug or add a feature (3 reports)
  • What was asked: read unfamiliar code, plan first, direct the coding agent, check its output, finish a working flow
  • Most common: No single repeated question

All 6 reports are from 2026. You work in the HackerRank IDE with a built-in coding agent. For 2026 interns it was the technical interview; one senior backend candidate had it next to a DSA round (Hire here, No Hire in DSA). One intern got a Java banking system and did not choose the language. A candidate who took it said the statement was vague: sketch the design yourself, then plan with the AI without drifting from your idea.

QuestionWhat it testsLevel
Banking codebase: allow only card PINs of 4 distinct digitsReading unfamiliar code, directing an AI agentIntern

System design interview

At a glance

  • Length: 60 minutes (Atlassian's guide and two reports)
  • Rounds: 1
  • What was asked: Atlassian-style products for experienced loops; chat, suggestion and phonebook systems in older campus loops
  • Most common: No design repeated

Atlassian's guide says you design on a whiteboard, paper or a drawing tool, from fundamentals upward, with attention to performance and reliability. One senior interviewer focused on REST endpoints. Two 2025 candidates were rejected or down-levelled after this round, one for shallow reasoning on choices such as GraphQL. No intern report had this round.

Reported designs.

  • A product management tool like Jira (SDE III, offer)
  • A tag management service for every Atlassian product: add, remove and search by tag
  • A spreadsheet service like Google Sheets, down to the schema and queries (2022 graduate)
  • A search suggestion system over millions of words: brute force, a trie, then caching
  • How WhatsApp works, then how push notifications reach a phone
  • A phonebook system, then how HDFS works

Values and management interviews

At a glance

  • Length: about 40 to 65 minutes in reports; Atlassian's guide says 45 for values and 60 for management
  • Questions: a few behavioral prompts, often built around one or two projects
  • Themes: conflict, planning, feedback, initiative, teamwork, Atlassian's five values, past impact
  • Most common: a conflict or disagreement and how you resolved it (5 reports)

Five campus and graduate reports had one combined M&V round; experienced loops split it into a hiring-manager round and a values round, and Atlassian says the values interviewer is usually from outside the team. These rounds are scored: one 2025 intern was rejected after the values round, and one P50 candidate was placed at P40 after the management round.

Reported questions.

  • A conflict or disagreement with a peer or lead, then how you persuaded them
  • Talk through Atlassian's values with your own examples
  • How you planned a big technical project
  • Your project and role, what you learned, and one mistake you won't repeat
  • A time you received feedback and how you handled it
  • A time you showed initiative and it had a positive effect
  • Successes, failures, and times you led versus took a back seat

See all behavioral questions in the question bank.

Stage 5: Hiring committee, team match and offer

Hiring committee. Atlassian's careers site says a Hiring Committee independent of the interviewers makes the decision from interview feedback and your CV. Levels can change: one 2025 candidate went from P50 to P40 and was then offered P30 after system design; another came out of the management round as a strong hire at P40 instead of P50; a third failed a P50 Karat screen and continued at P40. A 2026 senior with a DSA No Hire was still offered the design round.

Team match. Atlassian's backend guide says you then meet hiring managers with open roles on a Team Match call, which is not an interview, and have 24 hours to choose a team before the final offer.

Timing. Campus results came the same day in two drives and after a week in a third. A 2026 Canada intern was still waiting more than two weeks after positive feedback, and one 2026 candidate heard nothing for about four weeks after the values interview.

The Codinzhub Next-Round Playbook

The theme across the stages: finish working code, test it, then take the next requirement.

To get past shortlisting

  • Keep real projects on your resume and be ready to demo one; one intern thought students without projects were cut even after solving OA questions.

To get past the OA

  • Bank the easy question fast and spend the time on the DP or graph one. One graduate drive interviewed only students who solved all 3.
  • Partial credit counts: one 2024 intern passed 11 of 15 tests on the hard question and still cleared "at least 2.5 questions".

To get past the Karat or technical screen

  • Answer each design scenario in two or three minutes: the bottleneck, the fix and its trade-off. Four Karat reports describe 5 such questions.
  • Finish the first coding problem quickly; the second appears only if time allows. Three reports mention a redo, so use it if offered.

To get past the coding, code design and AI-enabled rounds

  • Code in your own IDE with a test framework ready, as Atlassian's guide asks, and run a test after each change.
  • Get the base running before designing for extensions; two P40 candidates fell short on a scale-up.
  • In the AI-enabled round, sketch the design before prompting and read every change the agent makes.

To get past system design

  • Clarify scope first, then define REST endpoints and the data model. Justify each technology; one down-level cited shallow reasoning about GraphQL.

To get past the values and management interviews

  • Prepare a story per value and per management area; one intern was rejected and one P50 candidate down-levelled here.

To get through the hiring committee and team match

  • Expect to wait: 2026 overseas and lateral candidates waited two to four weeks. Atlassian's guide gives 24 hours to choose a team.

How the loop differs by level

The building blocks are the same across levels, but campus loops are shorter, and experienced loops add a screen, a separate code design round and system design. Atlassian's ladder in these reports runs P30 (graduate), P40, P50 (senior) and P60 (principal).

LevelWhat candidates describe
Intern (15 reports)A 90-minute HackerRank OA with 2 to 4 questions (3 in most), one technical interview of 45 to 60 minutes (AI-enabled in 2026), then a values or M&V round. No system design. One 2024 intern was cut for not knowing ordered-map complexity.
New grad, P30 (6 reports)The same OA, one or two technical rounds, a design round in 4 of the 6 reports, then a grad values and management round. One 2026 off-campus graduate had 5 rounds, including live coding that was not DSA.
P40 (4 reports)Karat screen, Data Structures and Code Design rounds, then system design, management and values. Examples: snake game, lowest common ancestor. Two of the four were unsure after missing a scale-up.
P50 and senior (8 reports)Karat or a 45-minute engineer screen, two coding rounds (one AI-enabled in a 2026 report), system design, and separate management and values rounds. Two 2025 reports describe down-levelling to P40.

Campus interviews were packed into one or two days with quick results; lateral candidates moved stage by stage with a recruiter check after each.

Most-asked topics

Across 22 reports that named technical questions, the questions concentrated in hash-map-backed data structures and small classes, then intervals, strings and DP. Trees and graphs, often the centre of generic prep, tie for the smallest group here.

TopicWhat it meansSeen inReported examples
Hash maps and class-style data structuresStore and update state with maps, counters, lists and small classes11 of 22 reportsContent popularity tracker; LRU cache; agent rating system
Intervals, sorting, heaps and prefix sumsOrder events in time, keep the best item, sum ranges9 of 22 reportsMinimum rooms or platforms; employees in the office each hour; merge k sorted lists
Strings, sliding windows, stacks and two pointersScan text or arrays with a moving window or pointers8 of 22 reportsSmallest window with all characters; repeatedly delete string b from a; group strings with the same letters
Dynamic programming and mathsBuild answers from smaller answers; counting and number theory8 of 22 reportsFewest prime parts of a numeric string; most jobs with increasing values; no palindrome of length 5+
System design and data modellingDesign services, APIs and schemas, and reason about scale7 of 22 reportsTag management service; Jira-like tool; estimate server size
Trees and graphsNodes and edges: traversal, ancestors, connectivity7 of 22 reportsLowest common ancestor; last day you can still cross; minimum stress path

The question patterns that keep coming back

A recurring style in these reports wraps an ordinary technique in a product story: a channel's popularity, an agent's ratings, a file system, Confluence edits. Learn to see the pattern under the story.

1. Hash maps, counting and small class-style data structures

The largest group (11 of 22 technical reports) and the Atlassian signature: a small class with a few methods that must stay correct as requirements grow. How to choose: pair a hash map for lookup with a second structure that keeps the order you need: a linked list for recency (LRU), buckets or an ordered map for "most popular", a deque for a moving snake. The popularity tracker and the snake game were each reported twice.

Reported examples. Content popularity tracker; LRU cache; agent rating system with monthly averages; a rate limiter of two calls per second per customer; group strings made of the same letters.

2. Intervals, sorting, heaps and prefix sums

Seen in 9 of 22 reports, sometimes framed as time: meetings, trains, employees entering and leaving. How to choose: to count overlaps, sort the start and end points and sweep; to keep the best item while merging streams, use a heap; to fill a count for every hour, use a difference array and a prefix sum.

Reported examples. Minimum rooms or platforms; employees in the office each hour; merge k sorted lists, then with a multiset; a weighted job scheduling variant.

3. Strings, sliding windows, stacks and two pointers

Seen in 8 of 22 reports, mostly in OAs. How to choose: "shortest substring that meets a count" means a sliding window with a frequency map; "remove a pattern until none is left" means a stack; "pairs under a limit" after sorting means two pointers.

Reported examples. Smallest window with all characters; repeatedly delete string b from a; subarrays with product less than K; a Boats to Save People variation.

4. Dynamic programming and maths

Seen in 8 of 22 reports, half of them OAs. How to choose: if the answer for a prefix depends on a few earlier answers, write a prefix DP; if the rule is local (no palindrome of length 5 or more), keep only the last few characters as the state.

Reported examples. Fewest prime parts of a numeric string; most jobs with increasing values; fill '?' so no long palindrome; count arrays with an even sum.

5. System design and data modelling

Seen in 7 of 22 reports, from Karat scenarios to full designs, including Atlassian-like products. How to choose: start from users and operations, turn them into REST endpoints and a data model, then scale the hot path and say what fails first.

Reported examples. Tag management service; a Jira-like tool; a Google Sheets-like spreadsheet; estimate server size from daily users.

6. Trees and graphs

Seen in 7 of 22 reports. How to choose: for ancestors or paths, recurse and return what each subtree found; for "last day something is still possible", binary search on the day with BFS, or union-find in reverse; to count regions, DFS from each unvisited cell.

Reported examples. Lowest common ancestor; last day you can still cross; serialize and deserialize a binary tree; maximum path sum in a tree; count islands.

Preparing for the values interview

The values interview, and the management interview experienced candidates get beside it, is scored (see Stage 4). Atlassian lists five values: Open company, no bullshit; Build with heart and balance; Don't #@!% the customer; Play, as a team; and Be the change you seek. Its interview guide asks for real examples of living each one.

Build your story bank. Prepare six to eight real stories from projects, internships or work, each under two minutes and ending in a result. Cover the themes candidates were asked about:

ThemeReported prompt (paraphrased)Follow-up to expect
ConflictA disagreement with a peer, about how to code something, or with a project leadHow you persuaded them
ValuesSituations from your past that show Atlassian's valuesOne senior candidate covered all five
PlanningHow you planned a big technical projectNot reported
LearningYour project and your roleWhat you learned; one mistake you won't repeat
FeedbackA time you received feedbackHow you handled it
TeamworkCompromising your choices in a team; teammates who did not follow your decisionNot reported

Structure each story. Use STAR (situation, task, action, result), which two candidates said they used: say what you did, the result, and what you would change. Use real experiences only.

Expect depth. Campus rounds went deep on one project; one intern demoed a deployed project and was asked how they knew it served its purpose. Atlassian's backend guide says the management round covers driving outcomes, applying lessons, managing conflict and taking initiative.

A six-week plan

The plan follows the patterns above, most common first, and ends with timed mocks in your own IDE. Plan on two to three focused hours a day.

  1. Week 1: Hash maps, counting and class-style data structures. Popularity trackers, LRU cache, rating systems, rate limiters. Habit: write a unit test before each method.
  2. Week 2: Intervals, sorting, heaps and prefix sums. Meeting rooms with a heap and with a sweep line, merging sorted streams, difference arrays. Habit: state the cost of every structure, including ordered maps.
  3. Week 3: Strings, windows, stacks and two pointers, then DP. Windows with counts, stack reductions, prefix DP. Habit: say the brute force and its cost first.
  4. Week 4: Code design. Snake game, a file system with sizes, an agent rating system: 40 minutes to build, 20 to add your own scale-up. Habit: get the base running before you generalise.
  5. Week 5: Trees, graphs and system design. Lowest common ancestor, serialisation, binary search with BFS; Atlassian-style designs and Karat-style scenarios. Habit: start every design from users and endpoints.
  6. Week 6: Mocks. Four timed 60-minute mocks in your own IDE with two follow-ups and tests each, plus one set of five 4-minute design scenarios. 2026 interns: practise one change in an unfamiliar repository with an AI agent. Rehearse your values stories aloud.

Interns and campus graduates should weight weeks 1 to 3 for the OA; P40 and above should start code design and system design from week 2.

Only two weeks? Three days on hash maps and class-style problems, two on intervals and heaps, two on strings and DP, two on code design, one on trees and graphs, one on system design, and three on mocks and values stories.

Practice problems that match these experiences

These are well-known practice problems that match the patterns above. They are not claims about Atlassian's question bank unless a row says the problem was asked as-is.

#Practice problemPatternSeen in these experiences as
1All O`one Data StructureHash map plus ordered bucketsContent popularity tracker with O(1) most popular
2LRU CacheHash map plus doubly linked listAsked as-is in one campus report
3Group AnagramsHashing by a canonical keyGroup strings made of the same letters
4Minimum PlatformsSort and sweep over intervalsMinimum rooms or platforms so no intervals overlap
5Merge k Sorted ListsHeap, then balanced BSTAsked as-is in one SDE III report
6Remove All Occurrences of a SubstringStack-based string reductionAsked as-is in one campus OA
7Longest Increasing SubsequenceDP over a prefixMost jobs you can do when each value must exceed the last
8Search Suggestions SystemTrie prefix searchSearch suggestion system over a large dictionary
9Lowest Common Ancestor of a Binary TreeTree recursionLowest common ancestor in a tree you build first
10Last Day Where You Can Still CrossBinary search on the answer with BFS, or union-findAsked as-is in one SDE III report

How to handle a question you have never seen

Atlassian coding rounds are often one unfamiliar-looking problem that grows through follow-ups, so a repeatable method matters more than memorised solutions.

  1. Repeat the problem and one example. Ask about inputs, limits and edge cases before coding; one candidate's interviewer was unclear about the requirements, and clarifying questions took real time.
  2. Say the brute force and its complexity. Three selected campus candidates opened this way, and interviewers then asked them to optimise.
  3. Name the bottleneck. Repeated lookups suggest a hash map; repeated range counts suggest prefix sums; repeatedly needing the best item suggests a heap or an ordered map; "keep a moving shape or window" suggests a deque.
  4. Say when you are stuck. Interviewers gave hints in the reports (a hash map, a sweep line, BFS instead of DFS). One intern who took a hint still finished 10 minutes late, so ask early rather than late.
  5. Code in small pieces while talking. You have your own IDE, so write a class with clear method names, run it after each piece, and keep the base version working before the next requirement.
  6. Test one case and fix the wrong line. Write a unit test for the example and for each follow-up. One 2025 rejection cited an untested follow-up, and Atlassian's guide asks for tests.

Why candidates were rejected

Thirteen reports end in a rejection or a down-level, and two more describe failing one round while the process continued. Seven of the 13 give a reason. The reasons were rarely a missing algorithm and more often a slip under follow-ups.

  • Struggling in the data-structures round. Three reports: a 2023 senior candidate felt underprepared and stopped there; a 2026 senior got a DSA No Hire despite a Hire in the AI-enabled round; a 2025 P40 candidate got stuck on lowest common ancestor.
  • Correctness and testing gaps. Recruiter feedback to one 2025 candidate cited editing data inside a TreeSet without rebalancing and an untested follow-up, which the candidate disputed.
  • Weak complexity knowledge. One 2024 intern could not give an ordered map's insert and delete cost and felt that ended their run.
  • Searching for the algorithm. One 2025 candidate looked up the rate-limiting algorithm with permission and was not recommended.
  • Hints and time. One 2024 intern needed a hash-map hint, finished 10 minutes late and was not selected.
  • Depth in design or management. Recruiter feedback down-levelled one 2025 candidate for shallow reasoning on choices such as GraphQL; another dropped from P50 to P40 after the management round.

The six without a reason include rejections after a code-pair round, code design, system design and values; one candidate praised in the room was told feedback could not be shared.

Selected vs. rejected: do this, not that

No selected candidate described a perfect interview. A few habits repeated, and the rejected reports show what skipping them cost.

Do thisNot thatWhat the reports show
Test as you go: a quick test for the base and each follow-upUntested extensionsOne 2025 rejection cited an untested follow-up; Atlassian's guide calls testing crucial.
Brute force, then optimise, then edge casesA complex first design that eats the clockThree selected campus candidates used this order; one intern who started complex ran 10 minutes over.
Know your structure's costs: heap, ordered map, multisetUsing a structure you cannot analyseOne intern was cut over ordered-map complexity; one SDE III was asked how a multiset works.
Finish runnable code, then extendOver-engineering the first versionOne P40 candidate could not get runnable code for a scale-up; another ran out of time on an extensible design.
Treat values and management as scoredWinging the last roundOne intern was rejected after values; one P50 candidate was down-levelled after management.

If you remember only five things

  1. The process. Interns and campus graduates take a 90-minute HackerRank OA; experienced hires take a Karat or engineer screen; then come coding, code design or AI-enabled coding, system design from P40 up, and values and management rounds.
  2. The round format. A coding round is up to 60 minutes: one problem, often with follow-ups that extend your code.
  3. The core habit. Get working code running in your own IDE, test it, then extend it.
  4. The core topics. Hash-map-backed data structures and small class designs lead, then intervals and heaps, strings and windows, and DP.
  5. The last round counts. Values and management interviews rejected or down-levelled candidates; prepare real stories for the five values.

FAQ

Who takes the online assessment?

Interns and campus graduates do, and Atlassian's early-careers guide makes it mandatory for them. Experienced hires took a Karat screen or, in 2026, a 45-minute engineer screen instead.

Can I use AI during the interview?

Only in the AI-enabled round, which all six reports place in 2026. Three reports of other live coding rounds say AI was not allowed.

Is system design asked?

Yes from P40 up, as one 60-minute round. Four graduate campus reports also had one; no intern report did.

Which language should I use?

The one you know best; Atlassian says you choose. The exception was one 2026 intern's AI-enabled round, set in a Java codebase.

How long does it take?

Campus drives went from OA to interviews in one day to about a week, with results the same day or within a week. 2026 lateral and overseas candidates waited two to four weeks for decisions.

About this guide

The Codinzhub team researched and wrote this guide from 51 Atlassian Software Engineer interview reports, mostly from 2025 and 2026 with older campus reports for examples, including experiences and OA questions shared on Codinzhub, and checked the process against Atlassian's own interview guides. We kept only what candidates described. The questions are our own short summaries, and the practice links are similar problems unless marked as asked. If you interviewed recently, share your experience on Codinzhub and we will update this guide.

Disclaimer. Atlassian's process changes by role, year, level and office, and it changed noticeably in 2026. Use this guide to build a foundation, and confirm the current format with your recruiter.

Search the question bank

The questions from each round above, with practice links for similar problems. Log in to search every question we collected for this role, including questions shared on Codinzhub.

Showing 20 of 113 questions

  • Minimum stress path in an undirected weighted graph

    A Atlassian online assessment question shared on Codinzhub (IIT (BHU) Varanasi, intern).

    On CodinzhubIIT (BHU) Varanasi, internGraphsCodingOnline assessment
  • Maximum throughput after scaling servers connected in series

    A Atlassian online assessment question shared on Codinzhub (IIT (BHU) Varanasi, intern).

    On CodinzhubIIT (BHU) Varanasi, internNot reportedCodingOnline assessment
  • Lecture complexity array over a fixed schedule

    A Atlassian online assessment question shared on Codinzhub (IIIT Hyderabad, intern).

    On CodinzhubIIIT Hyderabad, internArraysCodingOnline assessment
  • Kth greatest element in every prefix of a permutation

    A Atlassian online assessment question shared on Codinzhub (IIT Roorkee).

    On CodinzhubIIT RoorkeeArraysCodingOnline assessment
  • Optimal team size for a hackathon from developer skill levels

    A Atlassian online assessment question shared on Codinzhub (IIT Roorkee).

    On CodinzhubIIT RoorkeeNot reportedCodingOnline assessment
  • Request-blocking system for IP addresses

    Level: Intern.

    Set and ordered-multiset bookkeepingCodingOnline assessment
  • Add and renew items with expiry times, then count unexpired items

    Level: Intern.

    Hash map simulation over timeCodingOnline assessment
  • Repeatedly delete every occurrence of string b from string a

    Level: Campus.

    Stack-based string reductionCodingOnline assessment
  • Smallest window that contains all characters of a string

    Level: Intern.

    Sliding window with countsCodingOnline assessment
  • Split a numeric string into the fewest prime parts

    Level: New grad.

    DP over prefixes with a sieveCodingOnline assessment
  • Given a slow or failing system, name the issue and fix it

    Level: P40.

    Caching, load balancing, fault toleranceCodingPhone or technical screen
  • Estimate server size from daily active users and growth

    Level: P40.

    Capacity estimationCodingPhone or technical screen
  • Model an order object with foreign keys and filter on them

    Level: P40.

    Relational data modellingCodingPhone or technical screen
  • Pros and cons of keeping data on the device versus the server

    Level: P40.

    Trade-off reasoningCodingPhone or technical screen
  • Minimum rooms or platforms so no intervals overlap

    Level: Intern.

    After a heap solution, redo it with a sweep lineCodingRound not stated
  • Content popularity tracker with O(1) increase, decrease and most popular

    Level: Intern.

    State insert and delete cost of an ordered mapCodingRound not stated
  • Merge k sorted lists, then with a multiset instead of a heap

    Level: SDE III.

    How a multiset and red-black tree workCodingRound not stated
  • Last day you can still cross a flooding grid

    Level: SDE III.

    Not reportedCodingRound not stated
  • Lowest common ancestor in a tree you first build from a diagram

    Level: P40.

    Not reportedCodingRound not stated
  • Longest streak with constant daily change, then allow one edit

    Level: P30.

    Change at most one elementCodingRound not stated
Page 1 of 6

Keep preparing on Codinzhub

Ready for the next topic?

Keep track of what you’ve learned.