Company interview guide
Atlassian Machine Learning Interview Guide (2026)
An ML craft deep dive first, then practical coding, ML system design for Jira and Confluence, and a values round that carries real weight.
Atlassian's machine learning interview loop has 8 reported stages, from recruiter call to debrief, hiring committee and team placement; typical length: Weeks to months.
Roles covered: ML Engineer, AI Engineer and Data Scientist. The question bank holds 16 questions reported from Atlassian interviews, 1 of them free to solve in your browser.
- Interview process length
- Weeks to months
- Stages
- 8
- Reports linked to bank questions
- 20
- Bank report dates
- Feb 2022–Mar 2026
- Questions in bank
- 16 (1 free)
- Last reviewed
- October 2026
How the Atlassian interview loop works
A recruiter call, then a first technical stage that eliminates many candidates: for ML engineers an ML craft round (project deep dive plus a short design), for data scientists a SQL screen. The full loop follows, typically two 60-minute coding rounds, an ML system design round, a hiring manager conversation and a values interview with someone from outside the team. Interviewers debrief and a hiring committee that only sees written feedback and your CV makes the call; team placement can come late. Candidates often report long gaps between stages, so follow up with your recruiter.
The Atlassian interview process, stage by stage
01Recruiter call
Short phone or video call
A conversation about your background, interests and the role, in which the recruiter usually lays out the rest of the process. Some candidates were told the team was not a fit at this point, and several report slow follow-up afterwards.
What it tests
- How your experience maps to the role and team
- Motivation for Atlassian and its products
- Logistics: level, location, timeline
How to prepare
- Ask which rounds you will face and whether the first technical round is an ML craft round or a coding screen.
- Ask how team placement works; some candidates are hired for a skill set and matched to a team later.
- Agree on a follow-up date, since gaps of several weeks are commonly reported.
A recruiter call opens the process in 13 candidate reports (5 ML, 8 data science; the stage is shared by every track). The company's hiring pages do not describe a recruiter call.
02ML craft round
About 60 min video call: roughly 30 min on one of your ML projects, 30 min on a small ML design
The first technical round for ML engineers and the main filter: many candidates stop here. You walk through a project you are proud of while the interviewer digs into choices and trade-offs, then design a small system or answer applied ML questions. Recent prompts included a recommender for Confluence articles, a simple RAG system across Jira and Confluence, a document classifier, and how to evaluate and speed up an LLM system.
What it tests
- Depth on a past ML project: problem framing, data, modelling choices, results, what failed
- Applied ML judgment: metrics such as precision versus recall for a given use case
- A compact ML or LLM design, often set in Atlassian products
- LLM evaluation offline and online, and serving latency (distillation, smaller models)
How to prepare
- Pick one project and rehearse a 10-minute version from problem to impact; start without waiting for prompts.
- Expect the design half to be outside your own specialty; clarify the goal and the label or relevance definition early.
- Know Atlassian's products well enough to ground a design in Jira issues or Confluence pages.
13 ML-role candidate reports since 2021 describe a first round built on a project deep dive, usually with a short design or applied ML questions; 2024-2026 reports call it ML craft. No company page describes it.
03Data science track: SQL screen, craft and case rounds
SQL screen (sometimes with Python), then a final round with advanced SQL, a business or product case, DS craft questions and a stakeholder interview
Data science candidates replace the ML craft and coding rounds with SQL and case work. Expect joins, aggregates and time-window questions on product data, then a case such as investigating a traffic drop or designing and reading an experiment. Some loops also ask ML basics such as regularization.
What it tests
- SQL across several tables: joins, aggregates, date windows
- Product metrics and root-cause reasoning for a business question
- A/B test design and interpretation
- Explaining analysis to stakeholders
How to prepare
- Drill funnel and conversion-within-N-days queries; they match reported questions.
- Frame cases around Atlassian's products and users; one candidate was rejected for lacking domain depth.
- Prepare a crisp experimentation framework: hypothesis, metric, power, guardrails, decision.
10 data scientist candidate reports since 2020 describe SQL screens followed by case, experimentation or DS craft rounds; formats vary by team. The company publishes no data science interview guide.
04Coding: data structures and code design
Usually 2 x 60 min live coding in your chosen language; one research applicant reported an online assessment first
Two practical coding rounds. The data structures round wraps a modest algorithm in a real-world scenario, such as assigning bookings to the fewest courts or a voting system. The code design round rewards clean, extensible code that grows with follow-ups, for example a streaming moving average that later becomes weighted, or a rate limiter. Candidates say the algorithms are not hard but the scenarios can be loosely specified, and one was rejected for leaving a follow-up unfinished.
What it tests
- Turning an ambiguous scenario into the right data structure (heaps, maps, sorted intervals)
- Clean, modular code that absorbs new requirements
- Streaming and windowed computation in constant memory
- Talking through trade-offs such as readability versus optimization
How to prepare
- Clarify inputs, ties and edge cases before coding; vague prompts are the main reported difficulty.
- Budget time so every follow-up gets an answer; incomplete follow-ups were cited in a rejection.
- Practise interval scheduling with a heap, sliding-window rate limiting and incremental averages until each takes under 20 minutes.
Free reported questions for this stage
- Moving Average Over a Data StreamCoding & AlgorithmsMediumSolve
9 ML-role candidate reports describe coding rounds, most since 2024 as data structures plus code design; the engineering handbook names both parts and lets you pick the language. One research applicant had an online assessment.
05ML system design
60 min design discussion
An end-to-end design, often an extension of the craft round and set in Atlassian's products. Reported prompts include a Confluence or content recommender, an internal chatbot that answers and acts over Jira and Confluence with permission checks, productionizing an LLM, and an ML labeling system. Some interviewers say the topic follows your resume, but candidates have received the team's own problem instead.
What it tests
- Framing the product goal and success metrics
- Retrieval, ranking and generation design for workplace content
- Access control, freshness and evaluation for LLM features
- Serving, monitoring and iteration plans
How to prepare
- Prepare one recommender and one permission-aware RAG design you can adapt to Jira or Confluence.
- State offline and online metrics for each component, including how you would evaluate answers.
- Expect follow-ups that change constraints; the handbook says design questions ladder up or down in difficulty.
9 ML-role candidate reports include a design round, most recently recommendation or RAG chatbot prompts; the engineering handbook describes a 60-minute systems design interview, not an ML-specific one.
06Hiring manager interview
One-on-one with the hiring manager or a senior manager
A conversation about your background, past projects and why they mattered to the business, plus how you work with others. Data science loops sometimes add a separate stakeholder interview, and some candidates met a manager who was not the hiring manager.
What it tests
- Business impact and reasoning behind past projects
- Collaboration and communication style
- Career goals and fit with the team's needs
How to prepare
- Explain the business case for each project you mention, not only the model.
- Have stories about working with senior leaders and resolving disagreements.
7 candidate reports since 2021 (3 ML, 4 data science; a behavioral stage shared by tracks) mention a hiring manager round; the engineering handbook describes a one-on-one with the hiring or a senior manager.
07Values interview
Conversational behavioral interview with someone outside your team
A behavioral round on Atlassian's five published values, run by an interviewer from another part of the company. Reported questions cover pushing back on a decision, driving change, mentoring, ambiguity, mistakes and conflict. Public prep guides warn that it can sink an otherwise strong technical loop.
What it tests
- Candour and openness with teammates and managers
- Customer focus in trade-offs
- Driving improvement and owning mistakes
- Team-first collaboration
How to prepare
- Map one or two STAR stories to each of the five values before the loop.
- Use concrete examples with your own actions and results; follow-up probing is common.
- Avoid ML jargon; the interviewer may come from sales, HR or support.
11 candidate reports (5 ML, 6 data science; shared by every track) describe a values round; the company's handbook and early careers guide also describe it, often run by someone from another function.
08Debrief, hiring committee and team placement
Internal review after the final round
Interviewers debrief, then a hiring committee that sees only written feedback and your CV makes the decision. One ML candidate was turned down at committee over a single unfinished coding follow-up. Some data science candidates describe hiring for a skill set first, with team matching or an open role settled later.
What it tests
- Consistency of signal across all rounds
- No single weak round, especially coding and values
How to prepare
- Treat every round as decisive; the committee reads written notes, not impressions.
- Ask your recruiter whether a specific team is attached to your offer.
One ML report mentions a hiring committee decision and two data science reports describe team matching after the loop; the engineering handbook describes an independent committee, but not team matching.
Questions and themes reported most often
- SQL on product data: joins, aggregates and within-N-days conversion (8 reports, data science track: sql screen, craft and case rounds)
- Walk through an ML project you are proud of, with deep follow-ups (7 reports, ml craft round)
- Product or experimentation case: metrics, A/B tests, a drop in traffic (5 reports, data science track: sql screen, craft and case rounds)
- Values stories: pushback, driving change, mistakes, conflict (4 reports, values interview)
- Recommender for Confluence articles or similar content (3 reports, ml system design)
- Moving average over a stream, then a weighted version without storing history (2 reports, coding: data structures and code design)
- Assign bookings to the fewest courts, then add maintenance gaps (2 reports, coding: data structures and code design)
- Rate limiter, as coding or as API design (2 reports, coding: data structures and code design)
What Atlassian asks about
- Coding: data structures & code design25%
- ML / LLM system design25%
- ML craft & project deep dive20%
- Values & behavioral15%
- SQL, product case & experimentation10%
- ML & LLM concepts5%
Atlassian interview prep checklist
- Rehearsed a 10-minute deep dive on one ML project, including what did not work
- Sketched a Confluence article recommender and a Jira plus Confluence RAG assistant with access control
- Coded a weighted streaming average in constant memory and an interval booking assigner with a heap
- Implemented a per-client sliding-window rate limiter with clean, extensible code
- Mapped STAR stories to all five Atlassian values
- Can explain how to evaluate an LLM feature offline and online and how to reduce its latency
Tips
- Treat the ML craft round as the gate: rehearse one project deep dive and one small design you can finish in 30 minutes.
- Learn the products: Jira issues, Confluence pages and the AI assistant work give context to most recent design prompts.
- Prepare a permission-aware RAG design and a content recommender; between them they cover most reported ML design prompts.
- For coding, practise realistic scenarios with follow-ups and finish every part rather than polishing one.
- Ask clarifying questions early; ambiguous prompts are the most common complaint in coding and design rounds.
- Write two STAR stories per company value; the values round is run by outsiders and can decide the outcome.
- Data scientists should drill multi-table SQL and experiment cases tied to a subscription software business.
- Keep momentum by asking your recruiter for timelines after each stage; slow follow-up is widely reported.
See the rest of the Atlassian loop
The later stages, the questions reported for them and how to prepare are in Pro.
Unlock with ProPractice the Atlassian loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
Atlassian interview FAQ
How many interview rounds does Atlassian have for ML roles?
Candidates report 8 stages: Recruiter call, ML craft round, Data science track: SQL screen, craft and case rounds, Coding: data structures and code design, ML system design, Hiring manager interview, Values interview and Debrief, hiring committee and team placement. A recruiter call, then a first technical stage that eliminates many candidates: for ML engineers an ML craft round (project deep dive plus a short design), for data scientists a SQL screen. The full loop follows, typically two 60-minute coding rounds, an ML system design round, a hiring manager conversation and a values interview with someone from outside the team. Interviewers debrief and a hiring committee that only sees written feedback and your CV makes the call; team placement can come late. Candidates often report long gaps between stages, so follow up with your recruiter.
How long does the Atlassian ML interview process take?
Reported timelines run Weeks to months (A few weeks to a few months).
What does Atlassian ask in ML interviews?
By reported question mix: Coding: data structures & code design 25%; ML / LLM system design 25%; ML craft & project deep dive 20%; Values & behavioral 15%; SQL, product case & experimentation 10%; ML & LLM concepts 5%. The most repeated themes: SQL on product data: joins, aggregates and within-N-days conversion; Walk through an ML project you are proud of, with deep follow-ups; Product or experimentation case: metrics, A/B tests, a drop in traffic.
How should I prepare for the Atlassian ML interview?
Treat the ML craft round as the gate: rehearse one project deep dive and one small design you can finish in 30 minutes. Learn the products: Jira issues, Confluence pages and the AI assistant work give context to most recent design prompts. Prepare a permission-aware RAG design and a content recommender; between them they cover most reported ML design prompts.
Are there free Atlassian interview questions?
Yes. LeetML has 16 questions reported from Atlassian interviews, and 1 of them are free to solve in the browser; they are listed by stage above.
Related interview guides
- GleanGraph-search coding, an AI-assisted build round and enterprise ML design, judged on a small set of reports.
- SalesforceA team-run loop that mixes plain coding, ML and LLM fundamentals, applied design and a deep look at your own work.
- MicrosoftProject deep dives in most rounds, quick ML and LLM fundamentals, standard coding, from-scratch ML code and GenAI design, in a loop shaped by the hiring team.
- StripePractical multi-part coding, a timed dataset-to-model round, an ML bug squash and payments-flavoured ML design.
- DatabricksPractical coding, LLM-centred ML design on data-platform problems, and a long, structured panel.
- CanvaAn AI-assisted applied ML problem, ML coding without AI, ML design and a values-based behavioral round.