Just ended recently Google 26NG SDE VO, all three rounds of interviews have been completed. Here I will share my real experience and questions with you in detail for the reference of students who are preparing for Google.
Overall feeling
Google's VO atmosphere is generally friendly, but the depth of investigation is not low. The three rounds are pure Behavioral, BQ + Coding, and pure Coding, which have a relatively balanced examination of communication expression, engineering thinking, and code implementation capabilities. The interviewers are all of high standard and will proactively guide and give reasonable tips.
Round 1: Pure Behavioral (45 minutes)
The interviewer was a very friendly Asian guy. At the beginning, he took the initiative to chat for seven or eight minutes (about the weather, what dramas he was watching recently, how many years he had been in the United States, etc.). There was no sense of pressure at all. After getting closer, he got to the point.
Main questions:
- Share a practical case in which you identified and solved technical risks (I talked about the unstable SLA of third-party services, which was solved through monitoring + stress testing + downgrade solution).
- When facing a particularly tight delivery deadline, how do you sort out priorities and ensure launch (I used RICE scoring + MVP splitting + daily stand-up meetings).
- The experience of receiving critical feedback on the code (the problem was directly pointed out by Senior in the PR, how I humbly accepted it and systematically improved the code readability, test coverage and comment specifications).
This round focuses on examining real experience + depth of thinking, and the STAR method is the best way to answer.
Round 2: BQ + Coding (45 minutes)
The interviewer is an Indian engineer with a strong English accent and a fast speaking speed, but his logic is clear. I asked two BQs first, and then went directly to Coding.
BQ part:
- Why do you want to work at Google?
- Is there any project that you feel you did particularly well and exceeded everyone's expectations? Specifically?
Coding part:
- Classic questions: Convert a binary search tree into a sorted circular doubly linked list. I used in-order traversal + maintaining the tail pointer to establish a two-way connection, and finally connected the end to the end to form a ring. Focus on dealing with edge cases such as empty trees and single nodes.
- Interval query questions: Given an array and multiple interval queries, determine whether each subarray satisfies the odd-even alternation of adjacent elements. I first preprocess an auxiliary array to record whether adjacent parities are different, and then answer each query with prefix sum O(1), overall O(n + q). The interviewer is quite satisfied with the pre-processing idea.
This round was a bit tiring because of the accent and continuous switching, but overall it went smoothly.
Round 3: Pure Coding (45 minutes)
The interviewer in this round was an ABC (American Born Chinese), and the atmosphere was the most relaxed. He almost didn’t ask BQ, and he started writing code after just a few words of greeting. He was very kind, and would take the initiative to ask for ideas and give natural prompts.
First question: Observe node values from the left and right perspectives of the binary tree
- Left: Return visible nodes from bottom to top
- Right: Return visible nodes from top to bottom
I used layer-order traversal, recording the leftmost and rightmost nodes in each layer, and the left perspective result was finally reversed. Note that the root node does not appear repeatedly.
The second question: The longest strictly increasing path length in a binary search tree. The path can start and end at any node. I used DFS + global variables to maintain the maximum value, and each node returns the longest downward increasing length starting from itself. Although it is BST, strict increment cannot make full use of the relationship between the left and right subtrees, and pure O(n) DFS passes. The interviewer asked if I could use the properties of BST to optimize. I said that I could add judgment but the benefits would not be big, and he agreed.
This round of writing was the most comfortable and the experience was great.
Other inspection questions
The interviewer also asked some basic data structure questions:
- The difference and usage scenarios between list and tuple in Python
- The concepts of min-heap/max-heap and how to use the heap to find the k-th largest/k-th smallest element
- The concept of dictionary tree (Trie) and its application in Google products (search completion, spelling check, etc.)
- The underlying principles and conflict resolution strategies of hash tables
Also, I also asked OOP design questions about Google cloud services (class hierarchy of Compute Engine, Cloud Storage, and BigQuery).
Don’t let one interruption cause you to miss out on your favorite offer
The autumn recruitment in North America is getting more and more popular. What really determines the offer is often not how many questions you answer, but whether you can perform stably in high-pressure OA and VO scenarios. Many people usually do it, but suddenly get stuck when coding, follow-up or system design comes to a limited time. One mistake may directly miss the top offer of $180k+.
ProgramHelp is provided is FAANG engineer-level practical training. The instructors come from major companies such as Google, Meta, and Amazon, and are familiar with the logic of real interviews. Regardless of OA or VO, all are provided by real people with pure handwriting assistance. The code style is natural, suitable for personal expression, and reduces risks.
