<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by oa_vo interview support on Medium]]></title>
        <description><![CDATA[Stories by oa_vo interview support on Medium]]></description>
        <link>https://medium.com/@caoxinyi945?source=rss-11ef14fd9ac8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*2G2HTsqIdrGCBLwvYsPRvw.jpeg</url>
            <title>Stories by oa_vo interview support on Medium</title>
            <link>https://medium.com/@caoxinyi945?source=rss-11ef14fd9ac8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 04:26:11 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@caoxinyi945/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[TikTok Data Engineer Interview | How did I pass quickly?]]></title>
            <link>https://medium.com/@caoxinyi945/tiktok-data-engineer-interview-how-did-i-pass-quickly-b98986089676?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/b98986089676</guid>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Sat, 04 Oct 2025 05:45:20 GMT</pubDate>
            <atom:updated>2025-10-04T05:45:20.577Z</atom:updated>
            <content:encoded><![CDATA[<p>Recently helped a trainee pass the <a href="https://www.tiktok.com/">TikTok</a> Data Engineer (DE) position in Bay Area, United States. The biggest feeling he got from the whole process is that the questions are simpler than expected, but at the same time they fit TikTok’s business scenarios very well, and he can see that the company pays more attention to whether the candidates really understand large-scale data processing, data modeling, and the ability to combine with business.</p><p>Compared with many companies that prefer to examine complex algorithmic questions, TikTok DE interview style is more inclined to engineering practice. In the VO interview, we provided voice assistance and thought process for the student, and his answers were very smooth, and the interviewer gave feedback in the end that “the answers were logical and clear, and the thought process was like someone who has done related systems”.</p><p>I’ll give you a complete breakdown of the process, questions, answer ideas, and details of our assists below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*36Bby_JnS90v7HvL.png" /></figure><h3>I. Interview process: accidental skip OA</h3><p>According to the process in HR’s email, the original plan was OA (online written test) + three rounds of VO interviews. But in reality, because of the recruitment scheduling problem, this participant directly skipped the OA and went to the three rounds of VO. This situation is not uncommon in TikTok, especially for the data engineer position with a more suitable background, sometimes it will be exempted from the written test, and go directly to the interview.</p><p>The three rounds of interviews are scheduled as follows:</p><ol><li><strong>First round: Hiring Manager Technical interview</strong></li></ol><ul><li><strong>Behavioral problems (BQ)</strong>: deep dive into past projects, especially Big Data and Data Warehouse experience; SQL questions: two questions, one on handwritten SQL output results and one on Hive script debugging; rhetorical questions session: about team, data scale, tech stack, etc.</li></ul><ol><li>The SQL part is not too difficult, it mainly examines whether candidates can reason logically to find out the execution order.The common error points in Hive questions are mismatch of field types, wrongly written partition fields, and sloppy syntax. In VO, we reminded the students to explain the SQL in a fixed order:FROM/JOIN → WHERE → GROUP BY → HAVING → ORDER BY, avoiding jumps in thought.</li><li><strong>Round 2: Easy Chat Round</strong><br>This round was surprisingly easy with almost no technical questions. The interviewer mainly talked about project experience, communication style, cross-team collaboration and career planning. The participant had prepared SQL and pipeline design, but it turned out to be like a coffee chat, in fact, the focus of this round was to confirm whether the candidate could fit into the team atmosphere.</li><li><strong>Round 3: Data Modeling (Data Modeling)</strong><br>This is the session that is closest to the actual work. The interviewer gave a business scenario: the tracking of short video playback and interaction metrics. Ask the participants:</li></ol><ul><li>Design table structures (Fact Tables / Dimension Tables); describe fields and granularity; explain scalability.</li></ul><ol><li>The interviewer even opened the HackerRank link, but in the end did not let him write SQL, but focused on schema design and logic. We remind the trainees during VO assistance that the order of answers must be Business Scenario → Fact Table → Dimension Table → Extensibility, so that the sense of structure is very clear.</li></ol><h3>II. Exclusive question sharing</h3><p>Although the overall difficulty is not high, the questions cover the three core directions of TikTok: large-scale data processing, recommendation system, and video storage architecture. Below is a summary of some of the questions and key points.</p><h3>1. Big Data Processing</h3><p><strong>Q1: How would you design a pipeline to process 100 billion video view events per day?</strong></p><ul><li>Data ingestion: Kafka</li><li>Real-time processing: Flink / Spark Streaming</li><li>Steps: cleaning invalid events → transformation (geo enrichment) → aggregation by user/video/region</li><li>Storage: ClickHouse / Druid for fast queries</li><li>Key points: exactly-once semantics, fault tolerance, scalability</li></ul><p><strong>Q2: How to detect trending videos in real time?</strong></p><ul><li>Define trending: growth rate of views, likes, shares</li><li>Sliding windows (5 min, 15 min, 1h)</li><li>Flink window aggregation</li><li>Store results in Redis for Top N queries</li></ul><p><strong>Q3: How do you handle Spark data skew?</strong></p><ul><li>Salting hot keys</li><li>Adaptive Query Execution (AQE)</li><li>Two-stage aggregation</li></ul><p><strong>Q4: How to model user behavior in a data warehouse?</strong></p><ul><li>Fact tables: video_views, likes, comments</li><li>Dimension tables: dim_user, dim_video, dim_time, dim_location</li><li>Consider granularity &amp; slowly changing dimensions (SCD)</li></ul><p><strong>Q5: SQL optimization techniques?</strong></p><ul><li>Use EXPLAIN to analyze query plan</li><li>Indexing, join optimization, early filtering, avoid full scans</li></ul><h3>2. Real-time Recommendation System</h3><p><strong>Q6: Design a real-time recommendation pipeline.</strong></p><ul><li>Event stream: clicks, watch time, swipes → Kafka</li><li>Real-time feature generation: Flink (user embedding, video popularity)</li><li>Feature storage: Redis/Tair</li><li>Model inference service → recommendation list</li><li>Online evaluation: A/B testing</li></ul><p><strong>Q7: What metrics do you use to evaluate recommendation models?</strong></p><ul><li>Offline: AUC, LogLoss</li><li>Online: CTR, CVR, watch time, diversity, novelty</li><li>Implementation: Kafka → Flink → dashboard</li></ul><h3>3. Video Data Architecture</h3><p><strong>Q11: How to store trillions of videos efficiently?</strong></p><ul><li>Video files: S3 / Ceph (object storage)</li><li>Metadata: Cassandra / HBase</li><li>Playback acceleration: CDN caching</li><li>Consider cost, latency, reliability, scalability</li></ul><p><strong>Q12: How to detect and remove duplicate or near-duplicate videos?</strong></p><ul><li>Exact: MD5/SHA-256 hash</li><li>Near-duplicate: perceptual hash (pHash, dHash), Hamming distance</li><li>Run async during upload pipeline</li></ul><h3>III. Summary of interviews</h3><p>As a whole, there are a few notable features of this TikTok DE interview:</p><ol><li><strong>The difficulty is on the watery side</strong>: Hardly any complex algorithm questions, more scenario and architecture questions;</li><li><strong>Exam point clarification</strong>: Big data processing, recommender systems, data warehouse modeling;</li><li><strong>Business Fit</strong>: All topics are highly relevant to TikTok’s short video business;</li><li><strong>pragmatic in style</strong>: More value is placed on the ability to get engineering on the ground than on paper.</li></ol><p>The final feedback from the trainees was:</p><blockquote><em>“During the VO process, with your voice reminders, I was able to quickly grasp the key points of my answers, and the interviewer was nodding throughout.”</em></blockquote><p>Overall, TikTok’s DE interview is more like a discussion of “business scenarios + data engineering practices” rather than a problem-solving competition. If you can answer from the path of business logic → technical architecture → scalability, you can basically get it.</p><h3>More than preparation, it’s a guarantee for the real world</h3><p>If you’re also preparing for a data engineer interview at TikTok or any other big player, you don’t have to do it alone. We <a href="https://programhelp.net/en/">Programhelp</a> can help you:</p><ul><li>OA Full ghostwriting to ensure 100% pass the test;</li><li>VO real-time voice assistance to help you answer the highlights at critical moments;</li><li>Interview frameworks are organized so that you are no longer nervous and your answers are clear.</li></ul><p>It is not difficult to get a big factory offer easily.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b98986089676" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Finished the Stripe OA + VO interview, The problems were too complicated]]></title>
            <link>https://medium.com/@caoxinyi945/finished-the-stripe-oa-vo-interview-the-problems-were-too-complicated-15ad90848320?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/15ad90848320</guid>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Tue, 23 Sep 2025 22:01:53 GMT</pubDate>
            <atom:updated>2025-09-23T22:01:53.568Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/840/1*9GLzyoRoq3Uusmf4KZGa8g.png" /><figcaption>Finished the Stripe OA + VO interview, The problems were too complicated</figcaption></figure><p><strong>Hackerrank:</strong> You have 60 minutes to solve one problem. It’s not like Leetcode, there are no complex algorithms, and it’s more about data processing. I passed all the test cases in 45 minutes. Two hours later, I received an invitation for a technical interview.</p><p><strong>Technical Interview:</strong> You have 60 minutes to solve a four-part problem. After completing each part, you’ll be given the next one, and the problems are related. I finished three parts, all passing the test cases. The last part was a bit tight on time, so I just wrote down the approach without testing it. The next day, I received an invitation for an on-site interview.</p><p><strong>VO Interview:</strong> There are three one-hour sessions. The overall interview process and problem types can be referred to here. The problem types are the same, and the SD is the original problem. I regret not checking it earlier…</p><ul><li>The first part is Coding. It’s basically a repeat of the phone interview, but this time I had to use built-in libraries. I only got two parts right. The interviewer seemed a bit disappointed because I wasn’t familiar with those libraries and had to look up a lot of things.</li><li>The second part is Bug Finding. To be honest, this interviewer was a bit annoying. First, you have to find the bug. He didn’t give me 30 seconds to think. If I paused for 5 seconds, he’d interrupt and tell me what to do next. Sometimes, when I was about to say what I was going to check next, he’d cut in and say exactly what I was about to say! I’m not sure if it’s his personality, but I hope it didn’t make me look bad. In the end, he gave me a small hint on where the bug was, and I fixed it immediately. Then, he added a small issue, which I also solved quickly.</li><li>The third part is Integration. It’s basically writing POST request code for an existing API, and there are many parts. I finished four parts, but time ran out. I’m not sure if they expect you to finish more, but I felt the most confident in this round. The interviewer didn’t interfere much, which was great for me because I need silence to think.</li></ul><p>The last round is the behavioral interview with HR. I’ll let you know the result if I have one.</p><p>By the way, my recruiter was super friendly and responsive. He’s a really good person.</p><p><a href="https://programhelp.net/en/">Programhelp</a> , The team consists of 7 engineers with backgrounds from top universities, including Oxford, Princeton, and Peking University, who have been working in Amazon, Google, Ali, and other top tier factories for many years.<br>We focus on OA writing, VO real-time assistance, interviews, covering written tests, remote interviews, algorithmic coaching, resume optimization, mock interviews and other full chain services. From online assessment Hackerrank package to FAANG interview assist and on behalf of the interview, to the whole escort straight to the Offer, we can provide you with efficient, seamless and safe support.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=15ad90848320" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Just finished the interview for Databricks, Share the interview questions]]></title>
            <link>https://medium.com/@caoxinyi945/just-finished-the-interview-for-databricks-share-the-interview-questions-2593ee2b863b?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/2593ee2b863b</guid>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Sun, 21 Sep 2025 16:33:02 GMT</pubDate>
            <atom:updated>2025-09-21T16:33:02.233Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/834/1*q6b6zQmEA3ju5XJw8smIbg.png" /><figcaption>Just finished the interview for Databricks, Share the interview questions</figcaption></figure><p>This was a <strong>VO interview with Databricks</strong>. Overall, the difficulty wasn’t too high, and it was divided into two parts: <strong>BQ</strong> (behavioral questions) and <strong>coding</strong>. The interviewer was an ABC lady, very easy to communicate with, and the whole atmosphere was relaxed without any pressure.</p><p>For the <strong>BQ part</strong>, as long as your answers are logical and demonstrate professionalism, it usually goes smoothly.</p><p>For the <strong>coding part</strong>, it mainly depends on your day-to-day experience and skills — though, of course, a little bit of luck helps too. If the problem isn’t too tough, the chances of passing are actually pretty good. Below is a brief walkthrough of my interview process:</p><h3>BQ:</h3><p><strong>Question:</strong></p><ol><li>Describe the most complex or challenging data-related project you have handled. What specific difficulties did you encounter, and how did you ultimately resolve them?</li><li><strong>Please share a major optimization project you worked on that significantly improved system performance, reliability, or scalability.</strong></li><li><strong>Can you tell me about an example where your work had a direct and significant positive impact on customers or end users?</strong></li><li><strong>Describe a project that you led end-to-end — from initial concept and technical design all the way through to final delivery.</strong></li></ol><h3>Coding</h3><h4>Code Question 1</h4><p>The problem allowed using Python’s time module to get the current time accurate to the second.</p><p><strong>Idea:</strong></p><p>We can maintain a deque (double-ended queue) to record the number of requests for each recent second. Every time there is an access:</p><p>If the current timestamp matches the one at the back of the queue, increment that second’s counter.</p><p>Otherwise, append a new record for the current second.</p><p>Meanwhile, we must continuously remove entries older than 300 seconds (5 minutes) to maintain the correct time window.</p><p>Finally, to compute the average access rate, simply sum all counts in the queue and divide by 300.</p><p><strong>Follow-up:</strong></p><ol><li>If the time span is very large but requests are sparse, how would you optimize memory usage?</li><li>If this system were deployed across multiple machines, how would you design it to track global access frequency?</li></ol><h4>Code Question 2</h4><p>This part mainly tests your solid technical skills, though a bit of luck also helps. The problem itself wasn’t very difficult — if you solve it properly, it’s quite easy to pass.</p><p>The prompt required using Python’s time library to obtain the current time in seconds. The solution idea was as follows:</p><ul><li>Maintain a double-ended queue to record the number of requests for each second.</li><li>On each request: if the current second matches the last timestamp in the queue, increment its counter; otherwise, append a new record.</li><li>Continuously remove entries older than 300 seconds to maintain the correct time window.</li><li>Finally, compute the average access rate by summing all counts and dividing by 300.</li></ul><p>There were also two <strong>follow-up questions</strong>:</p><ol><li>If the time span is very large but requests are sparse, how would you optimize memory usage?</li><li>If the system were deployed across multiple machines, how would you design it to track global access frequency?</li></ol><p><strong>Key points to note in the coding section:</strong></p><ul><li>Ensure proper removal of records older than 300 seconds.</li><li>Optimize the counting mechanism.</li><li>Avoid redundant queue operations.</li><li>Watch out for edge cases such as an empty queue or time rollbacks.</li></ul><h3>Want to get Databricks Offer quickly?</h3><p>ProgramHelp team provides <a href="https://programhelp.net/en/vo-help-hackerrank/">OA ghostwriting service</a>, interview assistance for a long time, Interview assistance services with guaranteed pass rate;</p><p>Remote voice assistance / real-time reminder, to help you in the interview process to avoid the “dig your own hole” situation.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2593ee2b863b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Meta E4 Interview Experience: Team Matching in Progress!]]></title>
            <link>https://medium.com/@caoxinyi945/meta-e4-interview-experience-team-matching-in-progress-72b994e47b3b?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/72b994e47b3b</guid>
            <category><![CDATA[job-search]]></category>
            <category><![CDATA[job-interview]]></category>
            <category><![CDATA[job-hunting]]></category>
            <category><![CDATA[meta]]></category>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Fri, 22 Aug 2025 07:20:50 GMT</pubDate>
            <atom:updated>2025-08-22T07:21:11.809Z</atom:updated>
            <content:encoded><![CDATA[<h3>Preparation</h3><p>I completed the first 150 LC problem list tags and marked the problems I found particularly difficult. The day before the interview, I briefly reviewed these approximately 30 problems. I scheduled the programming round approximately three weeks after the phone interview because I already had a foundation in the LC model and had completed the first 75 meta-problems required for the phone interview (preparation time: one week). For system design, I primarily read Grokking’s “Fundamentals of System Design Course.”</p><p>Then I mainly browsed Hellointerview. I first carefully read their design introduction and then did as many questions as possible on their website, focusing on topics discussed in LC. I also used their guided practice tools for hands-on experience.</p><h3>Round scheduling</h3><p>I tried to separate the rounds as much as possible so I could focus on coding and system design separately. I scheduled all coding rounds to be completed in one day, followed by system design and behavioral interviews a week later.</p><p>For coding rounds, I followed the same process for each question. First, I asked clarification questions, inquired about input types, discussed edge cases, discussed output types, and then outlined the solution approach. I made sure to visually demonstrate this to the interviewer. For example, I draw arrays and show pointers, or draw charts and illustrate the call stack for recursion. Then I briefly discuss the time complexity. I ask if I can write the code, then start writing it while verbally explaining my thought process, and finally run it with the interviewer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OgH-2t1vXQjWCa0fzMQVdQ.jpeg" /></figure><h3>The entire interview process</h3><h4>Round 1: Coding</h4><p><strong>Q1: Find the shortest time to reach the room</strong></p><p>For this question, I used knowledge related to Dijkstra’s Shortest Path for single-source shortest paths. I used a heap to store the positions to be visited, sorted them by time, popped the position with the least time each time, and then added the positions above, below, to the left, and to the right of that position to the heap.</p><p>I quickly found the optimal solution and demonstrated my approach to the interviewer. He deemed it the best solution, and I continued writing the code. During the simulated run, he mentioned a boundary case that needed testing. I conducted the test and found that the code already covered this scenario. He expressed satisfaction, and we proceeded to the next section.</p><p><strong>Q2: Numbers that appear only once</strong></p><p>My approach was to add the number twice the first time it appeared, ensuring the correct target number remained after two removals. The test passed smoothly, but it took a long time. Later, I discovered a faster solution on LeetCode: first use a set to remove duplicates (set(nums)), obtaining all unique numbers. All numbers appear three times, so calculate their total sum (3 * sum(set(nums))). Subtract the actual array’s total sum (sum(nums)), and the difference is 2 * the unique number. Divide by 2 to get the number that appears only once. After finishing, there were about 8 minutes left, so I had a pleasant conversation with him.</p><h4>Round 2: Coding</h4><p><strong>Q1: Merge two sorted arrays</strong></p><p>My approach to this problem was to merge the two already sorted arrays nums1 and nums2 into nums1, modifying nums1 in-place without using an additional array to store the result. Since nums1 had already reserved sufficient space (extra 0s), we could merge from back to front, avoiding the issue of data shifting caused by inserting elements at the front of nums1. I quickly solved this problem, and the interviewer fully understood my explanation and demonstration.</p><p><strong>Q2: Word Ladder</strong></p><p>My approach to this problem was to use BFS. First, traverse the list, and for each word, convert each position’s character to * and store it in a map. This means that if it’s a hit, there will be three entries: *it, h*t, and hi*. Then, use these as keys, with the values being lists of words containing these prefixes. This approach allows BFS to quickly find words differing by one character from the current word and add them to the queue. Of course, a map is not required; one could manually search each time, simplifying BFS by pushing words differing by one character into the queue while tracking the number of words searched. The interviewer was very impressed, as they had never seen another candidate solve the problem in this way and fully understood my approach.</p><h4>Round 3: Behavioral</h4><p><strong>Q: Standard behavioral questions, including conflicts, most proud of, and critical feedback.</strong></p><p>I felt I performed well in this round, following the STAR format and ensuring that the experiences I shared demonstrated leadership, growth, willingness to handle uncertainty, and teamwork. I made sure my self-reflection experiences involved learning. The interviewer and I had a great conversation for the last 8 minutes or so, and he told me I would love Meta because Meta places a strong emphasis on collaboration and offers many learning opportunities.</p><h4>Round 4: System Design</h4><p><strong>Q: Design Instagram</strong></p><p>I quickly completed the Hellointerview steps, including functional, non-functional, core entities, and API design. After confirming the focus with the interviewer, I moved on to the core content (high-level design). I completed the design and proposed a feasible solution. We discussed trade-offs, mentioned several potential issues and how to scale, answered all follow-up questions, and finalized the design. One drawback was that my entity design was not fully developed, primarily because the design focused on a specific section, and he assured me there was no need to worry about designing other parts. Therefore, I retained only the essential fields in the design and communicated this to him. We then had a productive exchange.</p><h3>Interview Summary &amp; Suggestions</h3><p>I was informed that I performed excellently in the programming and behavioral interview segments but only average in the SD segment. The recruiter mentioned that my application materials had passed the initial review and were being advanced to the candidate review (CR) or human resources (HC) stage, and requested that I participate in the subsequent SD segment, which I have scheduled for approximately two days later. I felt this performance was better than last time, as I could freely discuss all the topics I wanted to address and was quite satisfied with the final design proposal. I also answered all the interviewer’s questions.</p><p>Currently in the team matching phase! Though I haven’t received an offer yet, I’m already very happy to have reached this stage. If anyone encounters task difficulties while preparing for interviews, <a href="https://programhelp.net/vo/amazon-interview-questions-the-ultimate-guide/">feel free to come here for solutions!</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=72b994e47b3b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Amazon SDE 1 Interview Experience: Securing the Offer!]]></title>
            <link>https://medium.com/@caoxinyi945/amazon-sde-1-interview-experience-securing-the-offer-47a28b15afef?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/47a28b15afef</guid>
            <category><![CDATA[amazon]]></category>
            <category><![CDATA[job-hunting]]></category>
            <category><![CDATA[job-interview]]></category>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Thu, 21 Aug 2025 09:22:15 GMT</pubDate>
            <atom:updated>2025-08-21T09:22:33.101Z</atom:updated>
            <content:encoded><![CDATA[<h3>Timeline</h3><p>Application date: February 8</p><p>Received online assessment (OA) notification: February 13</p><p>Completed OA: February 14</p><p>First round of interviews: February 25</p><p>Second round of interviews: March 4</p><p>Third round of interviews: March 7 (rescheduled once)</p><p>Interview results notification: March 12</p><h3>OA</h3><p>Two DSA questions, to be completed within 70 minutes. Both questions are of medium difficulty, one involving strings and the other involving data structures. Due to participating in numerous online assessments and interviews recently, I do not recall the specific question content. It took approximately 25 minutes to complete.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a8LynvlcOSX3yzdbVIydow.jpeg" /></figure><h3>The Interview Process</h3><h4>Round 1：DSA</h4><p>This round of interviews was very rigorous and fast-paced. The interviewer joined Amazon three years ago and is currently an SDE Level 2. At the beginning of the interview, I introduced myself, and then the interviewer asked me three consecutive DSA questions, followed by a logic question.</p><p><strong>Q1: Task Scheduler</strong></p><p>My approach was to use an array to solve the task statistics, with tasks represented by uppercase letters in taskCount. I calculated the maximum number of times the task needs to be executed, then iterated through taskCount to obtain max. If max is calculated during task statistics, the number of calculations is the length of the tasks array. I calculated the number of tasks executed max times (maxCount), which can be combined with the second step. Based on the above results, the maximum value between (max — 1) * (n + 1) + maxCount and tasks.length is the shortest time.</p><p><strong>Q2: String multiplication</strong></p><p>Here, I considered the multiplication of numbers taught in elementary school, i.e., vertical multiplication. The first while loop clears the leading zeros, and the second while loop stores the integer array in the corresponding string. This problem can also be considered in the first for loop with carry, but that would add many judgment conditions, which I feel would be much more complicated than separating them.</p><p>Q3: Number of nodes in a complete binary tree</p><p>In a complete binary tree, except for the bottom layer which may not be fully filled, each layer contains the maximum number of nodes, and the nodes in the bottom layer are concentrated in the leftmost positions. If the bottom layer is the hth layer, it contains 1 to 2^h nodes.</p><p>I believe all three of these questions are of medium difficulty. The interviewer asked me multiple follow-up questions about my decision-making process, inquiring why I chose a specific data structure over other options, and requesting explanations for TC and SC, as well as why I believed a particular solution was superior.</p><p>With 4 minutes remaining, he asked me a question about LP, encouraging me to keep learning and maintaining curiosity. I don’t recall the exact question, but it was directly related to this topic, so I thought for 10 seconds and briefly shared an experience from one of my projects. Due to time constraints, there were no further questions about LP, and we concluded the interview in a relaxed atmosphere.</p><h4>Round 2：DSA</h4><p><strong>Q1: Planetary Collision</strong></p><p>This is actually a relatively straightforward problem. Imagine a large planet colliding with and shattering smaller planets moving toward it. Initially, I considered using a stack. However, to avoid the effort of manually transferring each element from the stack to an array at the end, I directly used the array as a stack instead.</p><p>I first explained the brute-force solution, followed by the optimized solution. Then, they asked me about my choice of data structure, time complexity (TC), and space complexity (SC).</p><p><strong>Q2: Minimum Sum of Descending Paths</strong></p><p>A descending path can start from any element in the first row and select one element from each row. The selected element in the next row can be at most one column away from the currently selected element (i.e., directly below or along the diagonal to the left or right).</p><p>I came up with a decent solution before time ran out. The interviewer asked me to explain my approach to solving the problem and to anticipate the TC and SC involved in that approach. I didn’t know how I came up with the optimal solution, but I did, and he was surprised that I could provide a solution so quickly. Overall, this round of the interview was better than the first round.</p><h4>Round 3：Behavioral</h4><p>First, the interviewer posed two logical questions:</p><p><strong>Q1: You are facing a very tight deadline in a project.</strong></p><p><strong>Q2: You have to make extra efforts to complete the project deliverables.</strong></p><p>I used the STAR format to implicitly answer these two questions. Follow-up questions involved certain decisions I made in these projects.</p><p>Then came the cross-examination. He asked numerous in-depth questions about my recent studies, followed by a detailed explanation and questioning of my graduation design project. The entire interview revolved around these topics. He inquired about the decisions made at each stage of the project, why those decisions were made, how conclusions were reached, which metrics were used, how those metrics were interpreted, the roles played by team members, and a significant amount of content related to project testing.</p><p>Since the project involved a lot of operating system and networking concepts, he also asked questions related to those topics, such as deadlocks, mutexes, race conditions, paging, segmentation, page replacement algorithms, jitter, semaphores, and other operating system knowledge. Networking-related questions included switches, routers, IMAP, SMTP, DHCP, hubs, firewalls, enterprise firewalls, subnets, proxy servers, VPNs, and tunneling technologies.</p><p>To be honest, I don’t remember all the details. He asked questions, and as I answered, he would follow up with counterquestions or quickly move on to another topic. Keeping up with the interviewer’s pace was extremely challenging during this round. I felt like crying because the questions became too difficult to answer, and I could only manage to respond using the scattered knowledge I had learned in college. However, overall, the learning experiences from my university days still proved useful.</p><h3>Interview Summary &amp; Suggestions</h3><p>I received the interview results on March 12th and was offered the position. I never expected to receive an offer after performing so poorly in the third round of interviews, but somehow, that’s how it turned out. Having a solid foundation of knowledge is crucial, as interviewers may ask a lot of follow-up questions during the interview. Without a strong knowledge base, it’s difficult to keep up with the interviewer’s pace. When preparing for an interview, it is essential to practice extensively and spend time on mock interviews. This will help you stay calm during the actual interview. If you encounter any issues, <a href="https://programhelp.net/vo/amazon-interview-questions-the-ultimate-guide/">feel free to come here for answers!</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=47a28b15afef" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Google SWE Four-Round Interview Experience: Rejected!]]></title>
            <link>https://medium.com/@caoxinyi945/google-swe-four-round-interview-experience-rejected-ad28675bd0a3?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/ad28675bd0a3</guid>
            <category><![CDATA[job-interview]]></category>
            <category><![CDATA[google]]></category>
            <category><![CDATA[job-search]]></category>
            <category><![CDATA[job-hunting]]></category>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Wed, 20 Aug 2025 07:57:52 GMT</pubDate>
            <atom:updated>2025-08-20T07:57:52.136Z</atom:updated>
            <content:encoded><![CDATA[<h3>Background</h3><p>I submitted my application in early May, and the recruiter first contacted me at the end of June to arrange a phone interview. The next morning, I received an email confirming that I had passed the phone screening, and the following day I spoke with the recruiter to arrange an on-site interview. I requested three weeks to prepare, and the on-site interview was scheduled for mid-August.</p><h3><strong>Phone Screen</strong></h3><p>I was asked a LC medium level question which involved dp and math related to prime numbers. The initial question was quite easy and i discussed my approach and wrote the code within 20 minutes, the follow up built on the previous question and need to cache previous results. I explained to solution and wrote down the code in the remaining time. The interviews asked me abouth the time complexties and if there were any optimizations that i could think off.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*f8K7O2UMYfjgjinlcWrS6g.jpeg" /></figure><h3>The Interview Process</h3><h4>Round 1：Googlyeness</h4><p><strong>The questions I was asked:</strong></p><p>Q1: If company morale is low and everyone is lacking motivation lately, what would you do?</p><p>Q2: How would you prioritize tasks when handling multiple projects simultaneously?</p><p>Q3: How would you manage to complete a complex project within a tight deadline?</p><p>The interviewer was very friendly and helped me relax during the interview. The questions primarily revolved around scenario-based scenarios related to teamwork, conflict resolution, and time management. I had a very in-depth conversation with the interviewer, which felt more like chatting with a colleague than a formal interview.</p><h4>Round 2：Coding</h4><p><strong>Q: Collecting Rainwater</strong></p><p>This is a classic two-pointer problem. First, it is important to understand that if the left and right pointers are to collect rainwater, as they move closer to the center, the maximum values on both sides must increase to ensure that the water in the middle can be collected. Only calculate the lower side of the current left and right pointers, as the higher side will inevitably cover it. The movement of the left and right pointers is determined by whether the calculation on that side is complete, meaning that only one side moves in each iteration.</p><p>My approach was to update the maximum height of both the left and right pointers in each iteration before they meet. To determine which side has the smaller value, calculate by adding the maximum value of one side minus the current height.</p><p>This problem is not difficult, but the interviewer posed four follow-up questions within 45 minutes. Each subsequent question was more challenging than the previous one, requiring the use of sliding windows, two pointers, and dynamic programming.</p><p>I successfully found the optimal solution and time complexity for four of the questions, but the interviewer posed the final question with only three minutes remaining, and I was unable to come up with a solution within the remaining time. Ultimately, he explained how to solve the problem, which involved using the solutions from the first three questions. Except for the final question, I believe I found optimal solutions and good code for all the questions asked. I felt I answered well, and the interviewer seemed satisfied.</p><h4>Round 3：Coding</h4><p><strong>Q: Climbing stairs</strong></p><p>The key to this question is to find the dynamic transfer equation, and most of them can be optimized in terms of space. For example, when climbing stairs, only the number of steps in the first two layers is retained, reducing the space complexity from O(n) to O(1).</p><p>I successfully explained the optimized solution to this problem, discussed the time complexity, and wrote the code (the code was quite large and took a long time to write). I was then asked follow-up questions.</p><p>I presented two different solutions to the interviewer, who nodded in approval. Ultimately, I chose one of them, explained its time complexity advantages and why it was superior to the other solution, and began writing the code.</p><p>During the coding process, with 5 minutes remaining, the interviewer suddenly stopped and mentioned that he typically reserves the final 5 minutes for candidates to ask questions. I asked him about his experience joining Google and the situation of his team, but only received brief and general answers. The interview concluded with 3 minutes remaining.</p><h4>Round 4：Coding</h4><p><strong>Q: Given a linked list, determine whether it contains a cycle.</strong></p><p>Here, I use fast and slow pointers. First, set two pointers: a slow pointer and a fast pointer. Then, start checking for cycles. The slow and fast pointers start from the head of the linked list. The fast pointer moves two steps at a time, while the slow pointer moves one step at a time. If the fast pointer reaches None (the end of the linked list), it means the linked list has no cycle, and we directly return None. If the fast and slow pointers meet at some point, it means there is a cycle in the linked list.</p><p>Reset the fast pointer back to the head of the linked list, while the slow pointer remains at the point where they met. Both pointers move at the same speed (one step at a time). When the slow and fast pointers meet again, the node where they meet is the starting point of the loop. I felt the interviewer was very satisfied with my performance, and this round of the interview was very enjoyable for me.</p><h3>Interview Summary &amp; Suggestions</h3><p>After my onsite interview, I waited a full three weeks for the recruiter to contact me, only to be told “no.” I felt I had performed quite well, aside from a few minor mistakes, but clearly, “quite well” wasn’t good enough for Google.</p><p>I hope this article helps those preparing for a Google interview. The questions themselves aren’t complicated, but they’re designed to be open-ended, and they expect you to ask a lot of clarifying questions. I believe these questions are intended to comprehensively test your foundational knowledge and problem-solving abilities. Although I felt optimistic after the interview, I think the mistake I made was failing to complete all the questions within the allotted time in both interviews. Although I explained the solution to one of the questions, I was unable to complete the code. If you have any questions about the interview process, feel free to message me. You can <a href="https://programhelp.net/vo/amazon-interview-questions-the-ultimate-guide/">come here to find solutions </a>for any difficulties or uncertainties you encounter during your preparation process~</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ad28675bd0a3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Uber SWE II (Software Engineer) Interview Experience: Let’s see how I got hired!]]></title>
            <link>https://medium.com/@caoxinyi945/uber-swe-ii-software-engineer-interview-experience-lets-see-how-i-got-hired-15a8b8f93868?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/15a8b8f93868</guid>
            <category><![CDATA[job-interview]]></category>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[uber]]></category>
            <category><![CDATA[job-hunting]]></category>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Tue, 19 Aug 2025 07:34:45 GMT</pubDate>
            <atom:updated>2025-08-19T07:34:45.687Z</atom:updated>
            <content:encoded><![CDATA[<h3>Job Application</h3><p>On August 4, I applied for the Software Engineer II Front-End position through the Uber Careers website and received an email on August 12 inviting me to take an online assessment (OA), which consisted of four Leetcode-style questions. I had one week to submit my answers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ap_oYEY6Zz3e6YnDRHOOLw.jpeg" /></figure><h3>OA</h3><p>The online assessment consisted of two simple-level questions on string manipulation and arrays, one difficult 2D dynamic programming question, and one moderately difficult tree-related question.</p><p>Within the first 40 minutes, I successfully passed all test cases for the first two questions but got stuck on the third question, so I decided to skip it. For the fourth question, I passed 70% of the test cases. Finally, before the deadline, I revisited the third question and used a brute-force N2 algorithm, passing 30% of the test cases. Due to time constraints, I submitted the OA.</p><p>I was very worried that my submission would fail, but the next day I received an email inviting me to schedule an interview with HR. I scheduled the interview for the following day.</p><h3>HR Interview</h3><p>The recruiter asked some common behavioral questions and technical questions about the tech stack, and provided detailed information about compensation and benefits at Uber. Later that day, they sent an email informing me that I had advanced to the next stage: the phone interview.</p><h3>Phone Interview</h3><p>One Leetcode intermediate-level 2D grid backtracking problem, followed by a Leetcode advanced-level problem. I successfully wrote the first version and completed most of the follow-up, but was unable to complete the subsequent versions within the specified time. Therefore, I explained how I would approach solving the problem if given more time. I was concerned that failing to complete the subsequent problems might lead to rejection, but ultimately I received an email confirming my success.</p><h3>Onsite Interview Preparation</h3><p>This time, I requested a month to prepare (they typically suggest two weeks, but if you need more time, I recommend trying to request it). For behavioral and leadership soft skills interviews, I watched Jeff H Sipe’s videos on YouTube and used the STAR method to document answers to the 15 most common questions I found online.</p><p>For LC question types, I worked on some problems from Neetcode 150 and Grind 75 (focusing mainly on arrays and hash tables) and searched online for common Uber questions, such as bus routes and merge intervals, and tried to solve them on my own.</p><p>When preparing for the front-end technical interview, I watched many YouTube videos about solving React interview questions and then tried to solve them on my own after watching the videos. I also searched Google for a list of common front-end questions and attempted to solve most of them on my own.</p><p>System Design Interviews: I watched over 20 mock interview videos and recorded all the content I found useful. Then I used the RADIO framework on Miro, the charting platform used by Uber, to independently solve some new problems.</p><h3>The Interview Process</h3><h4>Round 1：Behavioral</h4><p>BQs:</p><p>· Describe the most complex data science project you have worked on.</p><p>· How can data help Uber grow?</p><p>· What kind of work environment do you prefer?</p><p>I think this round of interviews went quite smoothly. The interviewers were very friendly and showed genuine interest in my past projects. Most of the questions they asked were ones I had prepared for in advance, and for those I hadn’t prepared for, I felt I handled them well.</p><p>Finally, they asked if I had any questions. I asked what the biggest challenge they had faced in their careers at Uber, and they said it was an excellent question and discussed it in detail. So, in an interview, building a good rapport with the interviewer is key. While there is an element of luck involved, it’s important to do your best to show that you are a worthy partner to work with.</p><h4>Round 2：DSA</h4><p><strong>Q1: Calculate the sum of two numbers</strong></p><p>This requires two loops, resulting in high time complexity. I used a hash table to solve this problem. The time complexity is O(n), using a dictionary data structure.</p><p><strong>Q2: Level Order Traversal of a Binary Tree</strong></p><p>My approach is to define two arrays, cur and nxt. The cur array stores the current node. During the traversal of the current node, its value is retrieved and stored separately in the array. The nxt array stores the child nodes of the current node. After processing the current node cur, the child nodes stored in nxt are assigned to the current node, and their values are added to the ans result. As long as the node is not empty, the previous operation is repeated.</p><h4>Round 3：Techinical</h4><p><strong>Q: Create a real-time chat interface (similar to Twitch or YouTube Live).</strong></p><p>The interviewer provided a function that simulates users receiving messages to help me solve the problem.</p><p>The problem focused on functional implementation rather than style design, so I used native HTML tags and avoided any fancy CSS. Subsequent questions involved understanding Promises, async/await, useEffect, useState, setTimeout, setInterval, and debounce/throttle, so I recommend thoroughly studying these concepts. I was able to complete all follow-up tasks, occasionally needing the interviewer to provide some hints.</p><h4>Round 4：System Design</h4><p><strong>Q: This question involves RADIO framework design.</strong></p><p>Since this was a front-end interview, I treated the server as a black box and did not consider load balancing, sharding, or databases. Then I was asked questions about rendering decisions, frameworks, communication protocols, and API design. Finally, they asked a few questions about security and stability, which I was unable to answer.</p><h3>Interview Process &amp; Suggestions</h3><p>A week later, I received an email from the recruiter inviting me to a 30-minute Zoom call. I was prepared to hear, “We’ve decided to continue considering other candidates,” but instead, I received a job offer. I got the job.</p><p>As you can see, my interview wasn’t perfect, so you don’t have to be flawless in every aspect to receive an offer from Uber. My advice is to focus on clear communication and ask clarifying questions before starting to code. Practice explaining your thought process as if you were doing a tutorial on YouTube. This way, the interviewers can follow your train of thought, and even if you get stuck, they can provide hints to help you continue solving the problem. If you remain silent, they won’t be able to assist you.</p><p>Feel free to ask questions anytime, and I’d love to hear about your experiences too! If you encounter any difficulties while preparing for interviews, <a href="https://programhelp.net/vo/amazon-interview-questions-the-ultimate-guide/">feel free to come here for solutions!</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=15a8b8f93868" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Google Four-Round Interview Experience: Ultimately Rejected!]]></title>
            <link>https://medium.com/@caoxinyi945/google-four-round-interview-experience-ultimately-rejected-ae565768787d?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/ae565768787d</guid>
            <category><![CDATA[google]]></category>
            <category><![CDATA[job-hunting]]></category>
            <category><![CDATA[new-grad]]></category>
            <category><![CDATA[job-search]]></category>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Mon, 18 Aug 2025 07:50:12 GMT</pubDate>
            <atom:updated>2025-08-18T07:50:12.281Z</atom:updated>
            <content:encoded><![CDATA[<h3>Background</h3><p>I recently had the privilege of participating in a Google interview, which was one of the most exciting and rewarding experiences of my life. Although I did not pass the final round of interviews, I learned a lot throughout the process and would like to share my experience with everyone.</p><p>When I received the rejection email, my good friend said something that left a deep impression on me: “Success never teaches as much as failure does.”</p><p>These words have guided me ever since, and I hope my story can inspire and help those who are preparing for their dream interviews or dealing with setbacks. Here is my experience:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/1*hjHSJR-Z_PF1ejgxSwjcrw.jpeg" /></figure><h3>The Interview Process</h3><h4>Round 1：DSA (1 hour)</h4><p><strong>Q: Given an undirected graph and a vertex in the graph, programmatically output the vertex sequences of the graph traversed using DFS/BFS before and after removing the given vertex.</strong></p><p><strong>Input Format</strong></p><p>First, input the number of vertices and the number of edges of the graph from standard input. The two integers are separated by a space. Then, starting from the next line, input the information for each edge on separate lines (each edge is represented by the vertex numbers at its two ends, separated by a space, the order of edge input and the order of vertex ID input for each edge can be arbitrary, but edges will not be repeated), and finally, the ID of the vertex to be deleted is entered on a new line.</p><p><strong>Output Format</strong></p><p>Output each vertex sequence in separate lines, with vertex numbers separated by a space. First, output the depth-first traversal vertex sequence and breadth-first traversal vertex sequence of vertex Q before deleting the given vertex, then output the depth-first traversal vertex sequence and breadth-first traversal vertex sequence of vertex Q after deleting the given vertex.</p><h4>Round 2：DSA (1 hour)</h4><p><strong>Q: Given a set of users and a log of friend relationships sorted by timestamp, find the earliest time when all users were first connected through friend relationships.</strong></p><p>Initially, each user belongs to an independent set. The number of sets at the beginning is n. My approach is to iterate through each log in chronological order, merging the sets containing two users at a time. I maintain the number of connected components, and when the count reaches 1, all users are connected. Since there is only one set left, all users are in the same set.</p><p>Termination condition: When all users are connected, immediately return the current timestamp.</p><h4>Round 3：45 mins DSA + 15 mins Googleyness</h4><p><strong>Q: Find the longest substring in a given string.</strong></p><p>First, confirm the specific details with the interviewer, such as whether it refers to a continuous sequence of characters, a substring containing all unique characters, or other criteria.</p><p>Follow up: Find the second-longest substring. If there are multiple substrings of the same length, please specify the expected output.</p><p><strong>Googleyness</strong></p><p>BQ1: How do you handle conflicts with teammates?</p><p>BQ2: How do you handle sudden changes in teams or projects?</p><p>Google behavioral interviews focus on values, adaptability, teamwork, and alignment with Google’s principles. Simply demonstrate your ability to reflect on personal experiences and leadership qualities, allowing the interviewer to see your capabilities.</p><h4>Round 4：45 mins DSA + 15 mins Googleyness</h4><p><strong>Q: How do I design an A/B test to verify whether a new recommendation algorithm can improve user engagement?</strong></p><p>My approach is to first define the primary metric, such as daily active time or click-through rate. Then determine the sample size using power analysis to calculate how many users are needed to detect a meaningful difference.</p><p>In terms of experimental design, I would use stratified randomization to ensure that the control and treatment groups are balanced across key dimensions, such as user activity and geographic location.</p><p>I would also consider the novelty effect and learning effect, which may require a ramp-up period to allow the algorithm to learn user preferences.</p><p>Finally, for result analysis, in addition to the primary metric, I would also examine secondary metrics such as user satisfaction and diversity to ensure there are no negative side effects.</p><p><strong>Googleyness</strong></p><p>The Google culture-related questions here prompted me to share how I tackle challenges, lead teams, and foster collaboration in the workplace. During this discussion, I felt a strong connection with Google’s core values.</p><h3>Post-Interview Reflections</h3><p>After completing all rounds, I spent a week reflecting on the entire process while anxiously awaiting the results. When I finally received the rejection notice, my feelings were mixed. Although I didn’t get the outcome I had hoped for, this experience was transformative.</p><h3>Interview Summary &amp; Suggestions</h3><p>Thorough preparation is key to passing an interview. Continuous practice in data structures and algorithms (DSA) will help you build a solid foundation. Mindset is crucial; viewing challenges as opportunities can help you maintain resilience and motivation. Mock interviews can enhance your problem-solving skills under time constraints. Constructive criticism is a stepping stone to improvement. Although I did not pass the Google interview, this experience was filled with growth and insights. I share my experience to inspire those preparing for interviews or facing setbacks. Let’s keep moving forward! If you encounter any difficulties or have any questions, feel free to <a href="https://programhelp.net/vo/amazon-interview-questions-the-ultimate-guide/">come here for solutions~</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ae565768787d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Amazon SDE 1 Interview Experience: I really got the offer!]]></title>
            <link>https://medium.com/@caoxinyi945/amazon-sde-1-interview-experience-i-really-got-the-offer-c792692f9248?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/c792692f9248</guid>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[interview]]></category>
            <category><![CDATA[job-interview]]></category>
            <category><![CDATA[job-hunting]]></category>
            <category><![CDATA[amazon]]></category>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Fri, 15 Aug 2025 07:03:20 GMT</pubDate>
            <atom:updated>2025-08-15T07:04:06.204Z</atom:updated>
            <content:encoded><![CDATA[<h3>Background</h3><p>5 years of undergrad software eng, 20 months of co op, mainly at a small company. Seems like this was only relevant for securing the OA/interview as the interviewers didn’t even have my resume pulled up. Finished OA with passing test cases 4 months ago (yes, 4 months). Preferred language is Python.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*snGHutfc4jvJEuzRtENRNQ.jpeg" /></figure><h3>The Interview Process</h3><h4>Round 1：LP</h4><p><strong>BQ 1: Tell me about a time when you received feedback from a senior or colleague and took action.</strong></p><p><strong>BQ 2: Tell me about a time when you needed to research a problem in depth.</strong></p><p><strong>BQ 3: Tell me about a time when you faced unfamiliar technology.</strong></p><p>I thought I had completely messed up. Although I knew about the STAR method, I had never practiced it. I had only jotted down a few story outlines in my notebook, but none of them were useful in this round. Therefore, my plan to follow the STAR method was completely unrealized, and I could only do my best to tell the stories I hadn’t prepared. The interviewers interrupted me several times and asked me to clarify, which I guess was because there were gaps in my stories. Perhaps this was a requirement of their process, which required more detailed questions. The only thing that gave me some comfort was that I didn’t completely freeze up and was able to provide answers, even without the STAR structure.</p><h4>Round 2：LLD</h4><p><strong>Q：Party playlist</strong></p><p><strong>My Approach：</strong></p><p>Use a hash map to count the relationship between songs and the number of people who like them. Use a set to store existing users. This way, adding and removing users is O1, and getNextSong is On.</p><p><strong>Follow up: If there are a lot of songs, how can getNextSong() be optimized?</strong></p><p>My idea here is to use a heap, but there is one rather complicated point about using a heap. It is not easy to delete items from a PriorityQueue in Java, as it requires traversal, so lazy update must be used. When calling getNextSong, a check is performed to see if the top element of the heap has expired, which results in an amortized time complexity of O1.</p><p>I think this was my best round, even though it took me longer than expected to finish. They didn’t pay much attention to the details of the implementation, and I actually made a lot of grammatical mistakes, such as performing subtraction between dictionaries and sets. But at least the concept behind my answer was correct, and I was able to explain through code how I arrived at multiple correct test cases, but I was corrected countless times during the process. Even with limited time, I still focused on writing clean code.</p><h4>Round 3：LP+LC</h4><p><strong>BQ 1: Tell me about a time when you were in a difficult situation.</strong></p><p><strong>BQ 2: Tell me about a time when you had to choose between a long-term decision and a short-term decision, and which one you ultimately chose.</strong></p><p>I referred to the prepared stories, but in this round, I didn’t follow the STAR method very well, nor did I focus on leadership principles. I just tried to tell these stories clearly so that the interviewer wouldn’t have to ask too many follow-up questions.</p><p><strong>LC: Implement methods for insertion, search, and string prefix matching. The search(word) method should return true if the word has already been inserted. The startsWith(word) method should return a list containing previously inserted strings that start with the word.</strong></p><p>I approached this LC problem through multiple iterations. I started with a list, then moved on to a dictionary, mentioned the time complexity of each method, and wrote readable code. The interviewer praised me for writing concise code. This was a smooth start, but things went downhill from there. After several awkward silences, I sensed that constructing some kind of tree structure might be the optimal solution. I brought this up, and fortunately, the interviewer nodded in agreement. I had not practiced Trie structures before. Here, I tried to mimic a linked list, but ultimately failed due to time constraints.</p><p><strong>Follow-up: If the Trie structure were correctly built, how would I handle it?</strong></p><p>Here, my answer was to use subtrees and postorder traversal to implement the startsWith method, which saved me in this round.</p><h3>Interview Summary &amp; Suggestions</h3><p>Within the next three days, I received the job offer! I still can’t believe it, not only because of the decision itself, but also because they sent the offer so quickly. My point isn’t that you won’t benefit from completing Neetcode 150 or spending a few days polishing and rehearsing your LP story, but I believe the factors that determine a candidate’s performance go far beyond what’s highlighted in the promotional materials. I believe my passion, clarity, and readability compensated for the lack of structure in the LP section and the inability to address the LC issues.</p><p>Finally, if anyone encounters any difficulties or has questions during their interview preparation, <a href="https://programhelp.net/vo/amazon-interview-questions-the-ultimate-guide/">feel free to come here for solutions!</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c792692f9248" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Amazon SDE Three-Round Interview Experience: Why Was I Rejected?]]></title>
            <link>https://medium.com/@caoxinyi945/amazon-sde-three-round-interview-experience-why-was-i-rejected-fa1785ca964c?source=rss-11ef14fd9ac8------2</link>
            <guid isPermaLink="false">https://medium.com/p/fa1785ca964c</guid>
            <category><![CDATA[jobs]]></category>
            <category><![CDATA[job-hunting]]></category>
            <category><![CDATA[job-interview]]></category>
            <category><![CDATA[new-grad]]></category>
            <category><![CDATA[amazon]]></category>
            <dc:creator><![CDATA[oa_vo interview support]]></dc:creator>
            <pubDate>Thu, 14 Aug 2025 07:17:13 GMT</pubDate>
            <atom:updated>2025-08-14T07:17:41.172Z</atom:updated>
            <content:encoded><![CDATA[<h3>Timeline</h3><p>· November 2024: Applied to Amazon.</p><p>· April 2025: Received a test email.</p><p>· May 2025: Received an interview invitation.</p><p>· June 2: Interview.</p><h3>Preparation</h3><p>I have been seriously preparing for Dynamic Number Theory (DSA) since December 2024. I have also studied graph theory, dynamic programming, and other topics, mainly through Striver’s videos, NEETCode 150, and Ashish’s Algomonster for practice.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OCWiovEuOaujlMBojnYNVQ.jpeg" /></figure><h3>The Interview Process</h3><h4>Round 1：Coding</h4><p>Q: Find the longest valid string.</p><p><strong>My Approach：</strong></p><p>Using the left starting pointer and the i pointer, you can construct a window and find the longest valid substring among all elements in the window. Elements within the window no longer need to be traversed and compared. When duplicate elements are encountered, simply assign left = index(c) — 1 to achieve a jump in left. Since the entire window spans from left to i, the length of the valid string can be calculated using i — left.</p><p><strong>Common mistakes:</strong></p><p>In the if statement, the longest length is calculated before updating left. At this point, right points to an invalid character. Outside the if statement, the longest length is right-left+1 for the valid part. Actually, calculating the length inside the statement is unnecessary. You only need to look at the valid part, otherwise it’s easy to get confused.</p><p>The interviewer asked me to start with a brute force solution and said we would discuss it further from there. I completed the problem using the brute-force method and asked many clarifying questions about the input and expected output. Overall, it was a good conversation, and I felt the interviewer was quite impressed with my problem-solving approach and conversational style. Finally, he explained the Trie, and I asked a few more questions. Although I didn’t use the Trie to solve the problem, I felt okay about it because I believe Amazon wouldn’t evaluate us based on data structures we’re unfamiliar with.</p><h4>Round 2：Coding</h4><p>Q: Pathfinding for social network friendships. Utilizing friend connections, design an algorithm to find the path from A to B and generate a list of related friends.</p><p><strong>My Approach：</strong></p><p>I used graph modeling to represent the friendship relationships, traversed the graph to find paths, and then output the path list. This is basically a basic graph traversal application, and either BFS or DFS can be used.</p><p>The interviewer was very friendly, and I felt quite confident after this round because I think the interviewer was also impressed.</p><h4>Round 3：Behavioral + Coding</h4><p><strong>LPs:</strong></p><p>Q1: Please briefly describe an experience you had when faced with unfamiliar technology.</p><p>Q2: Please describe an experience you had when you needed to research a problem in depth.</p><p>Q3: Please describe an experience you had when you disagreed with a colleague.</p><p>I had already familiarized myself with Amazon’s 16 leadership principles and prepared some relevant stories. During the interview, I basically followed the STAR method and applied some of my previous internship experiences.</p><p><strong>Coding:</strong></p><p>Q: Given an array of integers where no two elements are the same, find all pairs of numbers whose difference is equal to the “minimum absolute difference” and output them in ascending order (smaller numbers first).</p><p>My approach:</p><p>First sort the array (so that adjacent elements have the smallest difference), then perform a linear scan to find the smallest difference, and finally iterate through the array to collect all valid pairs.</p><p><strong>Follow-up</strong>: If the array contains a large number of elements (e.g., 10⁶), would sorting be too slow?</p><p>My answer: The problem states that the elements are distinct and integers, so counting sort could be used for optimization. However, since the input range is not specified, combining sorting with linear scanning is the most reliable approach.</p><h3>Summary of the Interview</h3><p>The day after the interview, I received a rejection email. I have no idea where I went wrong. For each data structures and algorithms (DSA) question, I had my own framework and didn’t jump straight into the solution. Instead, I asked clarifying questions and explained my thought process and approach along the way. During the third-round interview, the interviewer was very serious, which made me a bit nervous, but overall I was able to solve the problems and do my best to answer the linear programming questions.</p><p>I hope everyone will practice more and do more mock interviews when preparing for interviews. Be sure to take the interviewer’s suggestions seriously during the interview. Stick to your own problem-solving approach, but be careful not to be too stubborn, as that might lead to the same outcome as mine. If you encounter any difficulties or have any questions during the interview, feel free to <a href="https://programhelp.net/vo/amazon-interview-questions-the-ultimate-guide/">come here for answers~</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fa1785ca964c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>