Finished the interview Uber SDE New Grad interview, the overall process is quite typical. Black Chejia NG generally lasts for about 4 to 5 rounds, and the structure is relatively fixed: first OA, then a round of technical screening, and finally 3 rounds of onsite (mainly coding + behavioral). The whole process is very fast-paced and requires high algorithm proficiency and on-the-spot performance. This game has just been played, so while the memory is still fresh, let me give you a complete review and sort out the key test points and pitfalls.
Round 0: Filter Coding
Topic
The first round is a typical OA screening question. Given an integer array and limit, it is required to find the longest subarray length that satisfies the difference between any two elements in the subarray ≤ limit. The overall idea is very close to LeetCode 1438.
Problem-solving ideas
I advanced this question step by step according to the "optimization path": from violent enumeration of all sub-arrays (quadruple loop), optimization to triple loop to reduce repeated calculations, to dual pointers to maintain the maximum value within the window, and finally using sliding window + double-ended queue to optimize the time complexity to O(n).
The interviewer focuses on how you optimize step by step and whether you consider edge cases. At the same time, the analysis of time complexity and space complexity will also be more detailed. The overall performance was stable and passed smoothly.
Round 1: DSA + Set Union (DSU)
Topic
The topic of this round is the carpooling log problem: each record represents two users carpooling at a certain time. It is necessary to dynamically maintain the current number of vehicles and find out the final merge time point of the users in the same car.
Problem-solving ideas
At the beginning, BFS/DFS can be used to do connected components, but it is not efficient. After optimization, use union-find (DSU) to dynamically merge user sets through union operations while maintaining the number of connected blocks. The focus is on the implementation details of DSU: path compression, merge by rank, and the design of parent / rank / size. Also the complexity needs to be clearly stated (close to O(α(n))). The essence of this round is to examine the basic skills of data structure.
Round 2: Low-Level System Design (LLD)
Topic & Requirements
Designing a train-station management system requires supporting train scheduling, platform allocation, and time-based query capabilities, and requires writing runnable code.
Design ideas
I split the system into core classes such as Train, Platform, Scheduler, Manager, etc., used minHeap to manage the scheduling sequence, and designed a data structure that supports time dimension query. The overall design is biased toward scalability.
The interviewer will continue to challenge design decisions, such as why it is designed this way, whether there is a better solution, how to expand the system, etc. This round focuses more on engineering capabilities and code structure, rather than algorithms.
Round 3: High Level Design (HLD)
The fourth round is an upgraded version of system design, and the topic is to design the Uber Eats homepage.
The core test points focus on several directions: geographical location search (GeoHash / GeoIndex), API design, database selection (SQL vs NoSQL), caching strategy, and some typical indicator statistics (popular restaurants, popular dishes, order volume, etc.).
This round requires drawing a system architecture diagram and clearly explaining the responsibilities of each layer. The interviewer is generally friendly and will guide you step by step to optimize the solution, such as how to reduce latency and how to improve scalability, instead of denying your design right away.
Round 4: Manager + Behavioral
The last round is the manager interview, which is more behavioral. The overall atmosphere is relatively relaxed and mainly revolves around project experience. The focus is on several dimensions: code specifications, ownership (sense of responsibility), and whether what you do has actual business impact. It will also look at communication skills, such as whether you can explain complex issues clearly. Finally, there was time left for me to ask questions about the team. This round was more of a two-way choice.
Interview results & suggestions
This game has been played at a steady pace, and the final result is to get the Uber SDE Offer. If you are also preparing for Uber SDE NG, in fact, what many people are stuck on is not "not knowing how to do it", but stuck at critical moments: their ideas are stuck, their rhythm is messed up, or they start to doubt themselves in the middle of writing. In this case, having someone beside you to remind you of your thoughts in real time makes a big difference. Programhelp mainly handles high-intensity interview scenarios like this. Interview assistance Support, including OA/coding/system design.
If you are applying for a major company recently, or you happen to have an interview with Uber or a similar company soon, this kind of real-time assistance will be more direct and effective than simply answering questions, especially in the key rounds where "you can't make mistakes", the improvement is quite obvious.
