Just finished one recently Amazon VO Interview for 2026 SDE Intern. The whole process is two rounds of technical aspects, each round is a combination of Behavioral + Coding. The two interviewers are both Chinese, so communication is smoother and the rhythm is more natural. While my memory is still relatively clear, I will sort out the VO process this time and use it as a reference for students preparing for Amazon Intern interviews later.
Round 1
BQ session
The opening of the first round was fairly typical of Amazon Behavioral. The interviewer first asked a question about tight deadline. For example, how to deal with projects when time is very tight. After I gave the general background, the interviewer continued to ask for details.
For example:
- Who set this deadline?
- How much time was left for you?
- Are there other tasks happening at the same time?
- How do you prioritize tasks?
Basically, I asked this question for several rounds in a row. It felt like I wanted to confirm the authenticity of the matter and the logic of your decision-making at the time.
Then I was asked about a conflict experience with a colleague. This question was also asked very carefully, such as:
- What is the specific cause of the conflict?
- What were the views of both parties at that time?
- How did we finally reach an agreement?
There is also a more open question: Have you ever thought about what AI can be used for?
This question is more of a thinking type. There is no standard answer. It mainly depends on whether you usually have some technical thinking. The entire BQ session lasted about 30 minutes, and the level of in-depth exploration was relatively high.
Coding
The questions in the Coding part are not difficult overall. They are variations of LeetCode 215, but they only use an Amazon business scenario.
The core of the problem is to find the Kth largest element in a data set. Common solutions can use Heap or QuickSelect.
At that time, I chose the minimum heap solution. The overall idea is to maintain a heap of size k, continuously update the elements in the heap when traversing the array, and finally the top of the heap is the answer.
The process of implementing the code was relatively smooth, and no obvious stuck points were encountered. After completion, the interviewer asked me to briefly talk about the time complexity and space complexity, and then asked me to dry run the algorithm process with a simple example to confirm that there were no problems with the logic.
The first round of coding went relatively smoothly overall.
Round 2
The second round of interviews was scheduled on the same day, and the interviewer was still a Chinese man. The process is basically the same as the first round, with Behavioral first and then Coding.
Behavioral Interview
This round of Behavioral focuses on data-driven experiences.
The interviewer asked me to share an experience using data to make decisions. I first introduced the project background, data sources and analysis objectives, and then the interviewer began to ask more details.
For example, how the data was collected, what indicators were used to judge the results, and whether any abnormal data was encountered during the analysis process or the results did not meet expectations.
Next, the interviewer asked if any mistakes had been made during this process. For example, how the error occurred, how the problem was discovered, and how it was repaired later.
Finally, he further asked what improvements or new ideas there would be if we encounter a similar situation again in the future. This part was discussed very in-depth, and the entire Behavioral session probably lasted close to half an hour.
Coding
The second round of Coding is a question created by Amazon itself, not the original LeetCode question.
The core idea of the question is to use BFS (Breadth First Search). The whole problem is more like a state search problem, which requires starting from an initial state and expanding nodes layer by layer through BFS until a result that meets the conditions is found.
The implementation process mainly involves establishing a queue, then traversing the nodes layer by layer, generating new states at each layer and continuing to join the queue search. During the process, care needs to be taken to avoid repeated visits and at the same time handle some edge cases.
The overall difficulty of the question is not very high. After the code was written, the interviewer simply confirmed the logic and did not continue to ask follow up.
Interview results &
Finally, with the help of Programhelp, both rounds were successfully completed. If you are preparing for Amazon OA or VO recently, it is recommended to familiarize yourself with common question types in advance, such as Heap, BFS, Graph, etc., and prepare several sets of Behavioral stories to ensure that you can withstand continuous questioning. If you have no direction in your preparations, you can also refer to some real interviews. We have been sorting out OA and interview question banks from major companies such as Amazon, Google, Citadel, and Uber for a long time. We also provide real-time OA and interview questions. VO assist , giving reminders of ideas at key links to help successfully complete the entire process.
