Interviewers want to see how you break down information and navigate through problems. We also want to learn more about how you think, how you put yourself in the customer’s shoes, and how you make choices. Ask clarifying questions for any ambiguities and come up with a plan before you implement a solution. Managing your time is important, given the short 45-minute interview round.

We’re looking for engineers who can work not only with an existing platform, but who also understand what it takes to build or design a new system or platform from scratch.

Demonstrate best practices and ensure your code is clean, concise, and bug free. You’ll be asked to code only in the language you’re strong in. Don’t use pseudocode. You'll be using a third-party coding tool where you can run and compile your code. Practice, practice, practice. We can’t overstate this enough. This is a critical part of the interview.

At Microsoft, you test your own and other team members’ code. It’s likely that your interviewers will expect you to do some testing or show you understand how to test your solutions. When you finish your solution, don’t forget to test it before you say “done!” Remember, your basic responsibility is to make sure the feature you’ve coded works. Then, consider all the factors that could cause problems. What are the security implications of the feature? How can you stress this code? What are the boundaries and error conditions? Be sure to point out your corner cases.

Interviewers want to see what types of algorithms you use to implement your solutions. Recursion can be a common pattern for some types of questions. You should know the details of at least one n*log(n) sorting algorithm, preferably two (for example, quicksort and merge sort). You’ll need to explain the complexity of your algorithm (big O).

These are some common data structures that tend to come up often in questions: arrays and strings, queues and lists, linked lists, trees/tries, hash maps/hash sets, graphs. You should be familiar with when to use each one and describe their pros and cons.

Knowledge of distributed systems, SOA, and n-tiered software architecture is very important in answering system design questions. Resiliency, high availability, auto-scaling, replication, CAP theory, partitioning, all may be included. If you don’t work with these concepts regularly, make sure to review them before your interview.

We’ll be looking for your:

• Experience with machine learning models and training pipelines to perform model evaluation, profiling, and optimization. 

• Ability to apply, implement, and/or develop AI/ML algorithms and principles to solve complex problems.

You can expect questions about:

• Data Manipulation/Fluency/Data Insights and Exploration. Can you deal with data at various dataset levels, and how do you reason with the data as scale of data increases, run time, and so on.
• Probabilistic Thinking/Mathematics. There may be specific math or statistics questions and open-ended questions. We’re looking for how you process problems; how you analyze, clarify things, and draw solutions; and also how investigative you are about the data or results you have. 
• Statistics. Linear regression, time-series complexity, hypothesis testing/probability (t tests/p-values, and so on). 

Show your understanding of the layered model, application-specific protocols like HTTP, security-related protocols like TLS, network-level protocols like TCP and IP, and routing protocols like BGP and ISIS.