Microsoft Interview Questions
& Process for Software Engineers

Microsoft runs its software engineer interview as a single-day loop: a recruiter screen, an online assessment if your org sends one, then four or five 45-minute interviews back to back. What sets it apart from other big-tech loops is where the difficulty sits. The coding bar is easy to medium and leans classic, so the hard part moves elsewhere, above all to an object-oriented design round that below Senior takes the slot a distributed system design would fill anywhere else. Rather than sharding a datastore, you model an elevator in classes, or a parking lot, or a cache you extend into an LRU. The score comes from where you draw the class boundaries and what happens when the interviewer says “now add X.”

The loop belongs to the hiring team rather than to a central process, which is why so much of it flexes. Azure, Office, Windows, Gaming, and the Copilot orgs each decide whether you sit an assessment, how many rounds you get, and whether a distributed design round appears at all. Two things hold everywhere: the coding bar sits at easy to medium and leans classic, and behavioral questions open every interview instead of getting a slot of their own. Ask your recruiter which version of the loop you are in, because another candidate's Microsoft may share very little with yours.

The assessment runs on Codility or HackerRank, usually two or three problems inside 60 to 90 minutes, and it is graded by hidden test cases you never get to see. Whether you sit one at all depends on where you applied. Campus and university pipelines lean on it heavily, while some experienced-hire pipelines skip it and run a technical phone screen instead.

The bank it draws from barely overlaps the onsite. Its signature is a family of problems about building a string with no letter repeated three times in a row, which shows up under roughly five titles and is one greedy rule underneath. Max Network Rank and the minimum adjacent swaps to make a palindrome round out the core. Read the constraints before the prose, then write the empty string, the single character, and the all-identical input yourself, because the visible example leaves them out and the hidden suite does not. A solution that clears the sample and crashes on an empty string is scored as a fail.

Four or five interviews, about 45 minutes each, run back to back in one day and all virtual. Microsoft has no separate values round, so every one of them opens with ten or fifteen minutes of behavioral questions before anyone touches the editor. The mix is usually two or three coding rounds plus a design round, and the last slot on the schedule often belongs to the as-appropriate interviewer.

Coding Rounds

Two or three of them, one substantial problem each, at LeetCode easy to medium and leaning medium. The classics carry it: intervals, sliding window, trees, BFS and DFS across a grid, and the design-a-data-structure problems led by a Premium question. Genuine LC-hard turns up less than it does at Google or Meta, and the marks go to code that compiles, handles its edge cases, and reads like something a teammate could maintain.

Design

Below Senior this is usually the object-oriented round: an elevator, a parking lot, a vending machine, or a cache you extend into an LRU, modeled as classes and scored on how cleanly it absorbs the next requirement. At Senior (63 and up) and in Azure and infrastructure orgs it becomes a full distributed round, where the rate limiter is the top ask, ahead of Teams-style chat and OneDrive-style file sync.

The As-Appropriate Round

The last slot of the same day, run by a senior engineer or the hiring manager, and dropped often enough that a cancelled one tells you nothing. The name is a holdover from when that interviewer read the other write-ups first and decided whether the loop was worth continuing. The separate director-level veto round held weeks later, tidy pass percentages attached, is prep-blog folklore and does not appear in the loops candidates actually describe.

Behavioral

Not a round of its own. It is the first ten or fifteen minutes of every other round. Growth mindset is the centerpiece, the learn-it-all over the know-it-all, alongside respect, integrity, and accountability. The signature question is a failure and what you learned, and it only scores if you name the concrete thing you changed afterwards.

Microsoft's own careers guidance asks for clean, concise, bug-free code that you have actually tested, and that is very close to what the rooms grade. A correct, readable medium with its edge cases handled outscores a dense trick that saves three lines and hides an off-by-one. In a design round the same instinct scales up: real depth on the one component the interviewer probes beats a shallow tour of six boxes.

Every interviewer writes their feedback before anyone talks, and then the loop meets and decides. No committee reads your packet afterwards, and no trained bar-raiser sits in to hold a company-wide line, so the debrief is the end of it and the most senior person in the room carries the heaviest single vote. The behavioral signal accumulates across all four or five write-ups instead of living or dying in one. At Senior and in the Azure orgs, a thin design round is the most common reason an offer comes back a band below the title you interviewed for.

The ladder runs on internal numbers that candidates generally never see. You interview as an SDE II or a Senior SDE, and whether that meant a 61 or a 62 usually reaches you at or near the offer. Each title covers two numeric levels, and which of the pair you land on sets your band before anyone talks about money. Some recruiters now state the band up front and plenty still do not, so ask early and get the answer in writing.

Down-leveling is common here, and it is not always about you. A thin design round pulls a Senior target back to SDE II, and so does a headcount call that has nothing to do with your loop, which reaches you as “we are not recruiting Senior in this region right now.” Argue the level before you argue the money, since the gap between two bands dwarfs anything you can move inside one. The upside of a team-owned process is that a single no rarely ends things. Roles are team-specific, you can run loops with several teams in parallel, and strong candidates who lose out get routed to other teams that were interested.

Median total compensation per Levels.fyi (US, October 2026); self-reported and updated continuously.