← All posts
general

Microsoft Software Engineer Interview: Loop, Rounds, Prep

The Microsoft software engineer interview is harder to prepare for than Amazon's or Google's for one reason: there is no single process. Teams hire for themselves, the hiring manager picks the interviewers, and two candidates at the same level in Azure and in Office can sit quite different loops in the same week.

So the standard advice, grind medium problems and memorize a design template, covers about half of what you will meet. The other half is a domain round built around the team's product, a behavioral bar set by the company's culture attributes, and a final interviewer whose round only happens if the day has gone well. This guide covers each of those, with the caveat that most of it comes from candidate reports and third-party guides. Microsoft publishes very little, and what it does publish is quoted as such.

What Microsoft says and what candidates report

Microsoft's own interview tips page is short on structure. As of November 2026 it says that after the first conversation "the next interview steps will vary based on the role", that a recruiter will explain the format, and that the company aims to "assess fundamental knowledge, capability, and learning potential". It tells candidates to start a problem "by asking clarifying questions and stating your assumptions", and it recommends an extended STAR model for behavioral answers: Situation, Task, Action, Result, plus Reflection.

Everything else here is reported by candidates and prep sites, and they disagree in places. One guide puts the whole process at two weeks to a month; another at three to eight weeks. Treat the numbers below as ranges and ask your recruiter for the format of your own loop. They will usually tell you.

The stages in order

The sequence most candidates describe:

  1. Recruiter screen. Twenty to forty-five minutes on your background, the role and logistics. For senior roles the first call is sometimes with the hiring manager instead.
  2. Online assessment or technical phone screen. Some candidates get a timed coding test, often on Codility, sometimes on HackerRank, with one or two problems. Others go straight to a live screen in a shared editor.
  3. Hiring manager conversation. Reported for some teams, more often at senior levels. Conversational, about your past systems and what the team builds.
  4. The loop. Four to six interviews of 45 to 60 minutes, on one day or split across two, in person or on Teams.
  5. Decision, and sometimes team matching. Most candidates interview for one team. Some orgs run a shared loop and match afterwards.

Because teams run their own processes, candidates report being able to interview with more than one Microsoft team at once. If you are talking to two, tell both recruiters.

The loop round by round

A typical loop, according to guides built on conversations with Microsoft engineers, has two or three coding rounds, one system design round, one domain round and one behavioral round. Junior loops lean toward coding; senior loops trade a coding round for more design.

  • Coding. Reported as the round that counts most. One problem, sometimes two, in 45 to 60 minutes.
  • System design. From SDE II upward in most reports. At junior levels it is lighter or replaced by object-oriented design. Some SDE II candidates report a separate low-level design round, for example working out a design from a set of C++ function headers.
  • Domain. More common at Microsoft than at its peers. The questions come from the team's own area: cloud architecture for an Azure team, deployment risks for an AI team, API or database design for a backend role.
  • Behavioral. Rarely a whole round to itself. Expect it in the hiring manager's round and as the first ten minutes of the technical ones.

One guide reports that each interviewer in the final round scores the same four things regardless of the round's topic: collaboration across teams, adaptability, technical excellence and customer obsession. Whether or not every org uses that exact rubric, it matches what candidates describe. A coding interviewer at Microsoft is also writing down how you took a hint.

Guides disagree on how feedback is pooled. Some orgs collect written feedback asynchronously, some hold a live discussion, some do both, and grading scales differ by team.

The as-appropriate interviewer

The round Microsoft is known for is the last one. The "as appropriate" interviewer, AA or "as app" for short, is a senior person, usually the hiring manager's manager or a senior leader in the org, and by most accounts the person with the final say.

Three things candidates and guides report about it:

  • It is conditional. The name is literal: the round is scheduled as appropriate. If the earlier interviewers do not see enough signal, the day can end without it. Older accounts describe candidates who did poorly early being sent home without much feedback.
  • The interviewer has read the day. They see the earlier feedback before you walk in, and they often spend the round on whatever was weak or unclear: a design answer that stayed shallow, a behavioral story nobody quite believed.
  • Its content is not fixed. No guide describes a standard set of questions. Prepare for behavioral and technical follow-ups on what you said earlier in the day, and for the plain question of why this team.

A caveat. Two of the larger 2026 guides do not mention the round at all, and virtual loops spread over two days make "sent home early" harder to notice. Ask the recruiter whether your loop has an AA and who it is. If it does, reaching it is good news, but it is still an interview and candidates do fail it.

How to use it: before the last round, ask yourself which earlier answer you would redo. If the AA opens with the same topic, that is your second attempt. Take it plainly ("I thought about that design question over lunch; here is what I missed") instead of defending the first answer.

Coding at Microsoft

Candidates describe the coding rounds as conventional: medium-difficulty problems on arrays and strings, linked lists, trees and graphs, with tree questions the most often reported. Dynamic programming used to be rare and now shows up more, in its standard forms (coin change, edit distance, counting paths).

What differs from other large companies is the framing:

  • Questions tied to the team's work. Interviewers choose their own questions and often pick something close to what the team builds. A storage team asks about intervals and buffers; a compiler-adjacent team asks about trees.
  • The problem changes while you solve it. Several guides describe interviewers adding a constraint or altering the question mid-solution. That is a deliberate test of adaptability. Restructure calmly and say what the change costs.
  • Tested, explained code. Microsoft's own advice is to state assumptions and explain your reasoning. Interviewers reportedly weigh whether you walked through your code with an example before calling it done.
  • Tooling varies. Many teams use Codility for live rounds too. Ask in advance and spend half an hour in the editor before the day.

Language is your choice in most reports. If the team works in C++ or C#, expect a few questions about the language itself.

System design from senior up

At level 63 and above, guides say the design round carries the most weight and often decides the level. A strong coding day with a thin design round is the usual route to an offer one level down.

Reported prompts are product-shaped and often close to Microsoft's own business: design a chat feature for users of Azure, design a system that pushes software updates to a fleet of cars, with follow-ups on splitting large updates and handling lost packages.

Two themes come up more here than elsewhere:

  • Compliance and enterprise concerns. Auditing, logging, privacy, tenant isolation, data residency. Microsoft sells to enterprises and governments, and interviewers notice when a design has no answer for "who accessed this record and when".
  • Domain depth for senior and principal candidates. One guide says outright that a lack of domain knowledge at these levels can lead to down-leveling. If you are interviewing for an Azure networking team at 64, generic boxes and arrows will not carry the round, and a practice session with someone who has worked at Microsoft is the quickest way to find out how deep "deep" is.

The basics still apply: clarify requirements, put numbers on scale, name the trade-off you are making. The system design guide covers the general method, and the questions by level guide shows how the same prompt is graded differently for a senior and a principal engineer.

Behavioral themes

Microsoft does not have a list like Amazon's sixteen principles. Its interview tips page names the culture areas it cares about: growth mindset, diverse and inclusive, One Microsoft, customer obsession, values, and manager expectations for people managers.

In practice that means four kinds of question:

  • Growth mindset. A failure, critical feedback, something you had to learn fast. The interviewer wants what changed afterwards, which is why Microsoft adds Reflection to STAR.
  • Collaboration. Work that crossed team lines, shared credit, a disagreement with a partner team that ended in a decision.
  • Customer focus. How you found out what a customer needed, and a trade-off you made in their favor.
  • Adaptability. A priority that changed mid-project, an unclear goal, a reorg.

Have six to eight stories and know which theme each one serves. The companion post lists forty Microsoft behavioral questions by theme with what a strong answer contains.

Levels and what changes

Microsoft does not publish its level system. The mapping below is the one third-party guides and compensation sites agree on; titles in job postings vary.

Level Title What the loop looks for
59, 60 Software Engineer Correct, tested code on scoped problems; learning speed
61, 62 Software Engineer II Independence; owning a feature end to end; first real design round
63, 64 Senior Software Engineer Owning a system; design depth; direction for a team
65 to 67 Principal Software Engineer Architecture across an org; resolving cross-team technical disputes

Above 67 sit Partner and Distinguished Engineer, which are hired through a different conversation.

Two practical points. First, a level is a pair of numbers, and the offer names one of them: 63 and 64 are both Senior, with different pay bands. Ask the recruiter which number the role is opened at before the loop, and read the RSU guide before you discuss the offer. Second, the level is confirmed after the loop. If your design round reads as SDE II and the role is Senior, the offer follows the round.

Team-by-team variation

All of this varies by org, and candidates report real differences:

  • Azure and cloud teams build the domain round on cloud architecture, and guides give design prompts set inside Azure itself.
  • AI and Copilot teams reportedly add a round on machine learning for roles that touch it, with domain questions such as the risks of deploying a model.
  • Microsoft 365, Teams and gaming are named as hiring groups with their own domain questions; guides advise studying the product, its users and its common problems before the loop.

Since interviewers write their own questions, and some orgs use a rubric while others do not, the variance between two interviewers on the same team can be as large as the variance between orgs. The fix is information: ask the recruiter for the round list, the interviewers' roles and the editor. Then read the team's public engineering blog and product documentation, because the domain round is drawn from there.

A four-week plan

Week one: map the loop. Get the round list from the recruiter. Pick your language. Do ten medium problems on trees and graphs, out loud, each finished with a walked-through test case. Write down the team's product, its customers and its two or three obvious technical problems.

Week two: coding with interruptions. Fifteen more problems, adding linked lists, strings and standard dynamic programming. For half of them, have someone change a constraint at the twenty-minute mark. Draft six behavioral stories with a Reflection line each.

Week three: design and domain. Four design sessions: two general, two built on the team's product, each with an explicit pass on auditing, logging and tenant isolation. For the domain round, be ready to explain the last system you ran in production, including what broke.

Week four: full rehearsal. One mock loop: coding, design, behavioral, and a closing conversation with someone playing the AA who has read notes on the earlier rounds. That last part is the one you cannot rehearse alone, and it is where a session with one of the coaches who have worked at Microsoft helps most. Sleep before the loop; a five-round day punishes a tired fourth round.

Where coaching fits

The parts of this loop that reward a second person are the changing coding question, the domain round and the as-appropriate conversation. techinterview.coach lists six coaches with Microsoft on their history as of November 2026, and a mock session with a coach of that background can cover the round you are least sure of, with feedback on how your answers would read in the written notes. To rehearse the behavioral stories first, the AI practice interview runs a voice round and sends a written recap; it is free, sign in to start.

Related guides