Microsoft hails as one of the biggest cloud computing service providers with a team of several data scientists and data engineers. It mainly relies on Azure, AI tools, and Machine Learning. As many aspire to apply at Microsoft, here are some core competencies to master and tips to prepare for it.
About the Interview Process
The entire data scientist interview process consists of 3-4 tests (like online tests, written tests, technical tests, interviews, etc.) and takes about 2-4 days to complete. Candidates are shortlisted from the pool and taken to the next rounds based on scores obtained in each round.
Online Test: It is a CoCubes online test with multiple choice questions regarding basic machine learning, neural networks, deep learning, statistics, distribution, basic aptitude, logical reasoning, graph theory, etc.
Group Fly Round: It is an on-the-spot coding test. The candidates are provided coding questions that are supposed to be answered on pen and paper.
Interview Round: There are three interview rounds that the candidate has to undergo with various data scientists.
The online CoCubes test has around 15 questions to answer within 30 minutes. Results are announced within 2 hours, and candidates are shortlisted based on their scores.
- Brush up your basics, coding, computer networks, computation skills, etc.
- Go through as many competitive websites as possible and practice the basics.
- Do not entirely rely on last year’s trends.
- During the examination, check for negative markings as they may or may not be there.
Group Fly Round
Candidates are divided into groups of 3 or 4, and two coding questions are provided. They are supposed to write it down neatly with comments wherever possible. Then, the candidates are shortlisted for interview rounds.
Group Fly Round Interview Tips:
- Group fly round is about showcasing your approach. Before penning down your code, reach out to the invigilators around and tell them your thought process. They will guide you to think in the right direction and get to know your approach.
- Even if you cannot put forth the entire code, your method and approach will be noticed. It is one of the deciding factors for qualifying for the next round.
- Do not be lazy in putting comments.
- During preparation, go through as many coding questions as possible and solve them on your own (without looking at solutions unless necessary). Solve at least 2-3 every day.
- Words reversing, reversing numbers, etc.
- Matrix rotation of 1-D arrays.
- Finding unique function signatures with strings.
Five topics are broadly covered in round 1 interview: probability, statistics, data structures, algorithms, OOPS concept. Also, as the interviewer sees the resume, they might ask about the entire calculation and theory of why you picked up your past projects. Writing sample codes for various cases can also be asked.
It mostly consists of technical questions. The interviewer may ask questions ranging from deep learning, machine learning, past projects, SVM algorithms, kernels, various coding cases, decision tree algorithm, star schema, sampling methods (cluster, systematic, etc.), validation tests, etc.
Depending upon performance in each round, candidates are selected for the final HR interview.
It can be an HR or HR mixed technical round. The HR may focus on past projects, your contribution, and your learnings and ask for an explanation regarding implementation. Candidates can expect behavioural questions along with technical questions (coding, theory, etc.).
None of the three rounds lasts for more than 45-60 minutes.
Note: It is possible that instead of online tests and group fly rounds, a telephonic interview is conducted (called a phone screen), followed by a technical screen and onsite interview:
- The phone screen typically lasts for 30 minutes. The candidate is asked about his background, previous projects, and theory-based technical questions (based on machine learning or statistical problems).
- The technical screen lasts for 45-60 minutes, where the candidate is provided three coding questions (based on SQL coding, algorithms, probability, statistics, etc.). Here, the recruiters judge how well the candidate can solve the problems and explain their thought process convincingly.
- The onsite interview will last for a day at Microsoft, and various data scientists will conduct 3-5 rounds. Questions based on domain, behaviour, and codings can be asked.
- As various data scientists will interview the candidate at different rounds, you should be clear with your coding concepts.
- Say whatever you think because the interviewer wants to know your thought process (explain in-depth).
- Questions like a solution are already given; design the question to get that solution.
- Problem statements can be asked (like designing a fake news detection system), or situational problems can be given.
- Ask questions if you are not sure about a problem. The interviewer will guide you.
- Be prepared to answer these basic but critical questions in the HR round:
- Why do you want to work in Machine Learning?
- Why do you want to work with Microsoft?
- What do you expect from the job?
- Difference between lasso and ridge regression.
- What is a ROC Curve? Explain its meaning of sensitivity, specificity, and confusion matrix.
- OOPS concept.
- TRIE concept, benefits, and efficiency of implementation in data structures.
- BST loop and puzzle questions.
- Linked list, Floyd’s cycle detection algorithm, strings, etc.
Cracking Microsoft Data Scientist interviews is indeed a tough task and dream for many. The higher the rounds, the deeper and more difficult the questions get.
Hence, it is advisable to spend at least 2-6 months to clear your concepts and practice coding every day diligently to improve your problem-solving skills. Give mock interviews and solve lots of problems.
In the end, create a disciplined routine of studying and practising that you will follow daily. Stay motivated because, with lots of hard work and some luck, you will surely crack the interview.
You can also read our other blogs –