Recently I compiled a relatively complete set of Dropbox In summary, the SWE interview process and high-frequency questions are not the most technically difficult, but they have very high requirements for expression skills and cultural matching. They are a typical representative of "looking at comprehensive abilities." If you just answer the questions and don't prepare for the behavioral aspects, it's easy to overturn.
The following is based on the process + question type + core test points to help you sort it out at once.
Dropbox SWE interview process breakdown
OA
It is usually completed on HackerRank and usually contains 2 to 3 medium-difficulty algorithm questions, plus a small number of CS basic multiple-choice questions (operating system, network, data structure). The overall characteristic is that time is tight and coding proficiency is very high. The question itself is not particularly tricky, but many people get stuck "can't finish". This round is essentially screening basic abilities + hand speed. The core is whether you can quickly stabilize AC.
First round VO: Coding interview
The format is to write code online in real time, usually one algorithm question plus several follow-ups. The interviewer will continue to add conditions, such as space optimization, boundary processing, complexity analysis, etc. The focus of this round is not just on the questions themselves, but on whether you can express your ideas clearly, speak while writing, and adjust yourself when you get stuck. The overall approach is more of "communicative coding" rather than writing code silently.
Round 2: Technology Deep Digging
This round is easy to underestimate, but it is actually very critical. Content typically includes deep dives into resume projects (implementation details, trade-offs), small system design or API design, and data structure selection discussions. The interviewer pays more attention to whether you have actually done projects, whether you understand the reasons behind the design, and whether you have basic engineering thinking. The essence is not to test memorization, but to see how you make technical decisions.
Round 3: Behavioral aspect
Focusing on the company's values (Trust, Impact, Collaboration), we usually dig into 1 to 2 experiences and continue to ask for details. This round places a heavy emphasis on authenticity and consistency of expression, and many candidates will be distinguished here. It’s essentially the “decisive round” that determines whether you’re truly a good fit for the team.
Summary of high-frequency interview questions
Behavioral interview
- Self-introduction (focus: motivation + project)
- Conflict handling experience (teamwork)
- The most fulfilling project
- How to solve technical problems
- Have you ever made any innovative attempts?
Algorithm questions (LeetCode above average)
Frequent question types include sliding windows (e.g. Minimum window substring), hash + string processing (e.g. Finding duplicate files), greedy or DP (e.g. Maximum return on stocks). The overall characteristic is that I don’t take particularly biased questions, but I like to add follow-up very much, and I have high requirements for code quality and boundary processing. The core is not "whether it can be done", but "whether it can be done correctly stably and efficiently".
System Design/API Design
Common topics include designing file storage and synchronization systems, file sharing and permission control, and RESTful API design. Focus on system scalability, consistency and availability trade-offs, permissions and security mechanisms, and API specifications (such as version control and status code design). The overall focus is more on real engineering scenarios rather than purely theoretical design questions.
Machine learning (for related positions)
If it is an ML/DS position, it usually involves recommendation system design, user behavior modeling, model performance optimization, data cleaning and feature engineering, and hyperparameter tuning (such as grid search, random search, Bayesian optimization). The focus of the examination is on practical application capabilities rather than pure algorithmic derivation.
Dropbox SWE OA real question sharing
Coding
The first question is a standard algorithm question, and it is the original LeetCode question Candy (135), and the difficulty is Hard.
The general idea of the topic
There are n children in a row, each with a rating. Candies need to be distributed. Each child is required to have at least one candy, and the child with a higher score must get more candies than the adjacent children. The goal is to find the minimum number of candies needed.
Problem-solving ideas
If you pass this question, you will basically get points. The classic solution is to be greedy twice: first scan from left to right, then scan from right to left, and finally take the maximum value and add it up. If you haven’t seen it before, it’s still a bit stressful to think about it on the spot.
SQL (two lines)
Next are two SQL questions. The overall difficulty is not high, but the details are easy to make mistakes, especially time filtering.
The first question is to count the number of new user registrations in different countries. Given a db_user table, it is necessary to group by country and count the number of new users each month in the first quarter of 2021. The essence is GROUP BY + time range filtering.
The second question is to calculate user retention rate. Asked to find out how many users who registered in January 2021 were still active in February. Here you need to associate db_user with db_engagement, usually using LEFT JOIN, and then calculate the ratio.
These two questions mainly examine the basics of SQL, especially JOIN and date processing. If you are not familiar with it, it will be a waste of time.
Data/Pandas
The last question is a bit more data processing, and is done using Python Pandas. The question is given a CSV file containing customer data, and it is necessary to find the country with the largest number of contracts.
The trickier point is the tie-break rule: if multiple countries have the same number of contracts, you need to choose the country with a lower alphabetical order (for example, Z has priority over A).
The essence of the solution is to use groupby for aggregation, then sort_values, while paying attention to the details of the sorting rules, and finally output the results according to the specified format.
If you are pressed for time and want to pass the interview more safely
Many students often get stuck when preparing for Dropbox OA due to tight time and many hidden cases. If you are in a tight schedule, please feel free to contact me directly. Team members come from Oxford, Princeton, Peking University, Amazon, Google, and Alibaba. All OA ghostwriting, CodeSignal/HackerRank ghostwriting, VO assist All done by our own hands. Feel free to message me privately if you need anything, and we'll help you get ashore safely.
