Interview100x

Interview prep

How to prepare for a Microsoft software engineer interview

Microsoft publishes a technical interviewing guide that names four areas its engineering interviews evaluate, problem solving, design, coding and testing, and lists what to review for each, from recursion and O(n log n) sorts to replication, CAP and partitioning3. It asks for runnable code rather than pseudocode, and for testing that goes past the happy path into security, stress and boundary cases3. This page turns that guide into a prep plan and is plain about what Microsoft does not publish.

Last reviewed November 12, 2026

How the interview process works

Engineering candidates go through a loop of two to four conversations of up to about an hour1; the technical guide describes a short 45-minute technical round and what it evaluates3.
  1. 1.Application review and optional screen

    Applications are checked against the posted qualifications, and some roles add a short screening conversation before formal interviews1. For engineering roles, the posting's language and systems requirements are the first filter, so make sure your resume names them.

  2. 2.The interview loop

    Two to four conversations, each up to about an hour, by phone, on Microsoft Teams or in person, with potential teammates and cross-functional colleagues; engineering roles may require writing code1.

  3. 3.Technical rounds

    The guide describes a short 45-minute round format3, while the hiring page says conversations last up to about an hour1, so confirm the length. Each round can test problem solving, design, coding and testing, plus questions about your past work and scenarios from the role3.

  4. 4.Coding exercise or work sample

    Some roles add a coding exercise or work samples, and the recruiter explains the format and timeline for your loop12. Ask early whether yours runs in a tool that compiles, since the guide expects runnable code3.

What this is based on

Microsoft publishes a company-wide hiring page and a technical interviewing guide that lists evaluation areas and study topics13. It does not publish how many coding rounds a loop has at each level, how formats differ by team, or a separate new-graduate process (its student guidance page was not reachable when we checked). The one current posting used here is from the Microsoft AI organization and is not typical of every engineering team4.

Practice questions in the style of this role

Generated by Interview100x's question engine for this role. These are not questions reported from Microsoft.
  1. This role asks for bachelor's degree in computer science or related field, or equivalent experience. Where have you done that most recently?
  2. Walk me through a time you debugged a hard issue in a production service, including the language you used, the constraints you faced and what you did yourself.
  3. Tell me about a habit you rely on when working with teammates on a production service.
  4. Walk me through a time you profiled and tuned a service against latency or throughput targets.
  5. Walk me through how you found the bottleneck in a production service that was missing its latency or throughput target.
  6. Tell me about a key reliability, latency or throughput trade-off you decided on for a production service.
  7. Walk me through a hard production problem in a service where you had to narrow down the cause from limited evidence.
  8. Describe a time you worked with other engineers to resolve a disagreement about scope or quality on a production system.

A worked example

One practice question, a sample answer and what the evaluator said about it.

The question

Walk me through how you found the bottleneck in a production service that was missing its latency or throughput target.

Sample answer written for this demo

We had an internal API that returned a user's settings and permissions, and it was supposed to be under 200 milliseconds at p99, but it had crept up to something like 600 or 700. The first thing we did was look at the dashboards, and latency was worse at peak times, so we figured it was load-related. We looked at CPU on the hosts and it was fine, maybe 40 percent, so it was not compute. Then we added more tracing around the downstream calls, and we found that one call to the permissions database was taking most of the time. It turned out the query was doing a join that had gotten slow as the table grew, and there was also connection pool contention at peak, because the pool size was set really low from when the service was smaller. So we fixed both. We added an index, raised the pool size, and also added a small cache for the permissions lookups. After that latency came back down to well under target. It was a good lesson in not assuming the problem is where you think it is, and in having tracing in place before you need it.

Interview100x's evaluator output, unedited (fill-in gaps shown in brackets)

Ready

You identified two real causes of the slowdown: a join that got slower as the table grew, and connection pool contention at peak. Your account names the problems clearly, and the next step is to show how you measured them and what you changed.

  • What worked: You described two specific bottlenecks, a slow join and an undersized connection pool, and tied the pool problem to the service's earlier scale, which shows you understand how the system behaved.From the answer: It turned out the query was doing a join that had gotten slow as the table grew, and there was also connection pool contention at peak, because the pool size was set really low from when the service was smaller.
  • Next step: Name the latency or throughput target the service had to meet and what you measured with profiling to find each bottleneck.
  • Next step: Explain how you confirmed the join was the slow part and why the pool size change or query change worked, in terms an engineer would accept.
  • Next step: Pick one production performance problem you worked on and practice walking through the target, the profiling you did, the bottleneck evidence, the change you made, and how the system behaved under load.
  • Note: The slowdown had two causes. The query was doing a join that got slower as the table grew, and at peak there was connection pool contention, because the pool size was still set for the service's earlier, smaller load. I profiled the query and the pool usage to confirm each one, and then I fill in: [what change you made and what it did to latency or throughput].From the answer: It turned out the query was doing a join that had gotten slow as the table grew, and there was also connection pool contention at peak, because the pool size was set really low from when the service was smaller.

What the role is evaluated on

Problem solving
Microsoft says it watches how you break a problem down, consider the customer, clarify ambiguity and plan before implementing3. Good looks like restating the problem, asking about input size and constraints, and saying your plan out loud before you type.
Design
The guide covers two kinds of design: working within an existing platform and designing a new system from scratch3. If you do not use them daily, Microsoft suggests reviewing service-oriented and n-tier architecture, high availability, replication, CAP and partitioning3.
Coding
Microsoft asks for clean, concise, bug-free code in a language you know well, written in a tool that can compile and run code rather than pseudocode3. Practice in an editor that runs your code, not on paper.
Testing
Microsoft lists testing as its own area: first confirm the feature works, then cover security, stress, boundary, error and corner cases3. Say your test cases before you run anything.
Data structure choice and trade-offs
Microsoft's guide lists the core data structures to know, arrays, strings, queues, linked lists, trees and tries, hash maps and sets, and graphs, together with their trade-offs3. Good looks like saying why you picked a trie over a hash set, or a heap over sorting, and what it costs in time and memory, before you write the code.

Common ways answers fall short

  • Coding before clarifying

    A fast start on a misunderstood problem wastes the round even if the code is good. Microsoft counts clarifying ambiguity and planning as part of problem solving3, so spend the first minutes on questions and a short plan.

  • Testing only that the feature works

    Microsoft's guide treats testing as verifying the feature works and then considering security, stress, boundary, error and corner cases3. Before you call code finished, say how it behaves with hostile input and under heavy load, then walk through empty input, one element, duplicates and invalid values.

  • Interviewing in a language you only half know

    Choosing a language because it seems preferred, then fighting its syntax, costs time and confidence. Microsoft asks for a language you know well3. Use the one you can write without looking things up.

  • Design as box-drawing

    Drawing services and arrows without numbers, failure handling or trade-offs reads as shallow. For the existing-platform variant Microsoft describes3, the hard part is respecting constraints you did not choose. State load, what fails first and what you would give up.

  • Past-work stories with no owner

    Technical rounds can include questions about your past work3. Stories told entirely as 'we' leave the interviewer unsure what you did. Name your decisions, the alternatives you rejected and what you would change.

  • Rusty recursion and sorting

    Microsoft's guide specifically suggests reviewing recursion and at least one, ideally two, O(n log n) sorting algorithms such as quicksort and merge sort, with Big O analysis3. Writing merge sort with an off-by-one in the merge step, or being unable to state its space cost, is an avoidable loss on material the guide names.

A prep plan for the time you have

One week

Cover the four published areas once and test yourself under time

  1. Day 1: Reread your posting, mark which of problem solving, design, coding and testing it leans on, and ask your recruiter for your loop format.
  2. Days 1-2: Review Big O and recursion, then implement merge sort and quicksort from memory in your interview language.
  3. Days 2-4: Solve two timed problems a day in an editor that compiles, saying your plan first and listing test cases before you run the code.
  4. Day 5: Sketch one new system and one change to an existing system; for each, write two trade-offs and one failure scenario.
  5. Day 6: Write three STAR(R) stories about past technical work, each ending with what you would do differently.
  6. Day 7: Run one full timed practice session and spend the rest of the day on the weakest area it exposed.

Suggested pass: Trial Pass, $12One full session shows whether your narration and test coverage hold up under follow-up questions before you commit to more practice.

Two weeks

Build depth in data structures and design, then rehearse a full loop

  1. Week 1: Work through each data structure Microsoft lists (arrays, strings, queues, linked lists, trees and tries, hash maps and sets, graphs) with two problems each, noting why you chose it.
  2. Week 1: For every problem, write boundary, error and stress cases before running the code.
  3. Week 1: Review one distributed-systems idea a day: replication, partitioning, CAP, high availability, n-tier architecture.
  4. Week 2: Do two design sessions, one greenfield system and one extension of an existing service with latency targets.
  5. Week 2: Prepare four STAR(R) stories: feedback you applied, work across teams, a time you adapted, and a bug you owned.
  6. Week 2: Run two practice rounds back to back to simulate a multi-conversation day, then review where your explanations got vague.

Suggested pass: Sprint Pass, $99Daily rounds across coding, design and past-work questions for one job is what Sprint covers.

Four weeks

Diagnose, close the two biggest gaps, then rehearse the whole loop twice

  1. Week 1: Diagnose with one timed problem per data structure family and one design question, then rank your gaps.
  2. Week 2: Spend most practice time on your bottom two gaps and keep one mixed problem a day for everything else.
  3. Week 3: Alternate design days (existing platform, then new system) with coding days, and add a security or concurrency case to every test list.
  4. Week 3: if your posting is infrastructure work, prepare one performance problem you debugged in production: what you measured, the cause and the fix4.
  5. Week 4: Run two full mock loops on separate days, each with coding, design and a past-work conversation.
  6. Week 4: prepare questions for each interviewer about the team's systems, how they test them and how on-call works.

Suggested pass: Sprint Pass, $99Four weeks of preparation for one Microsoft loop fits Sprint; if you are interviewing at several companies at once, Search covers multiple jobs.

Questions candidates ask

How many interviews are in a Microsoft software engineer loop?
Microsoft says most loops are two to four conversations of up to about an hour each1. It does not publish a fixed number for engineering roles, so ask your recruiter, who explains the format for your specific loop1.
Can I write pseudocode in a Microsoft coding interview?
Microsoft's technical guide asks for code written in a tool that can compile and run it, rather than pseudocode, in a language you know well3. Practice the same way.
Which programming language should I use?
Microsoft asks for a language you know well and does not require a specific one in its guide3. The posting we reviewed lists Python or C++ as examples4, but needs vary by team, so check your own posting.
What algorithms and data structures does Microsoft say to review?
Its technical guide lists recursion, at least one and ideally two O(n log n) sorts such as quicksort and merge sort, Big O analysis, and core data structures with their trade-offs: arrays, strings, queues, linked lists, trees and tries, hash maps and sets, and graphs3.
What should I review for Microsoft system design questions?
Microsoft's guide suggests reviewing service-oriented architecture, n-tier design, high availability, replication, CAP and partitioning if you do not use them daily, and it covers both extending an existing platform and designing a new system3.

Practice with an interviewer that follows up

A Trial Pass is one full practice session with feedback on your own answers. It counts toward a Sprint Pass if you continue.

Practice this question now

Sources

  1. 1.How we hire | Microsoft Careers · careers.microsoft.com · undated · accessed 2026-11-12
  2. 2.Interview tips | Microsoft Careers · careers.microsoft.com · undated · accessed 2026-11-12
  3. 3.Technical interviewing | Microsoft Careers · careers.microsoft.com · undated · accessed 2026-11-12
  4. 4.Software Engineer, Model Serving System | Microsoft AI · microsoft.ai · undated · accessed 2026-11-12

Interview100x is not affiliated with, endorsed by, or sponsored by Microsoft. Company names are used only to identify the employer. Process details are summarized from Microsoft's public pages, linked above; they can change, so check the company's own careers site.

Last reviewed November 12, 2026