Hello,

Giving back to the community. Finished Amazon onsite for L5. There were a total of 4 rounds: 1 system design, 2 coding system design round, and 1 pure LC round. ALL ROUNDS BEGAN WITH 15 MINUTES OF LP.

1. System Design

   • Q: Design a top 50 movies recommender system for Netflix
     • Started off with requirements (both functional/non-functional)
       • Got the exact features we were trying to build, services that already exists or if I needed to build it
       • Talked about how many users will be using this, RPS, metrics to measure
     • Dived into what it would look like for customers, then transitioned to what it would look like for us
     • High-level Design
       • Talked about the flow of data (the second a customer clicks login into Netflix)
       • APIs we might need
       • Database to store our data (both for top 50 movies and user metrics)
     • Low-level Design
       • Dove deep into the APIs we were needing. Talked about what the implementation would look like (parameters, requests, etc)
       • Talked a lot about the database and how we know it's top 50
         • Dove down into primary/secondary designs, as well as cache
         • Talked about read/write ratio, how to update for this particular scenario
       • Talked about what would be needed to scale this upwards, where are the bottlenecks, and what happens during failures
   • Overall: Felt good about this round. Could've gotten deeper into the data-flow, but I think I was running low on time due to the 20minutes of LP at the beginning. I wanted to get all of my thoughts out, but should've just been more patient. I definitely felt like this round is similar to on the job.
   • If I had to score myself: 3.5/5

2. Coding System Design

   • Q: Design & code a scheduling system to perform jobs and their dependencies
     • Talked about the constraints/assumptions about this system.
     • Talked about what were the use cases and inputs/outputs
     • Designed a very high-level scheduling system that would take a job, dependencies and schedule a time for it.
       • Utilized a queue to keep track of the jobs in the pipeline
       • Utilized a hashmap to keep track of dependencies of a job before it can occur
       • One to thing to mention here is, the dependency can also have it's own dependency
   • Follow-Up: How do we do this across multiple systems? We shouldn't limit a single job to a single node performing our dependencies (i.e if we had 3 total nodes to utilize, and a job had 3 dependencies, each dependency should run on available nodes to reduce total time waiting)
   • Overall: Felt really good. Had great rapport with the interviewer. I really felt like it was us together trying to tackle this challenge. I definitely felt like this round resembles on the job.
   • If I had to score myself: 4.5/5

3. Coding System Design

   • Design & code a hub system (hub 1..n) that a robot can go to. A robot can go either forward or backwards m steps.
     • Talked about constraints and assumptions about this system
       • For example, let's say we have 5 hubs. Robot is at hub 4. We tell robot to move 3 hubs forward. Robot should be at hub 2.
     • Talked about what was the use cases and the inputs/outputs
     • Designed a high level circular linked list that can have the robots go indefinitely in circles depending on m steps
       • We shouldn't always move m steps if m steps is greater than hub. Use modular to get the actual diff.
       • We need to be able to dynamically add new hubs (either at the front, end, OR after a given node).
       • We need to be able to dynamically remove hubs
       • Utilized a simple circular linked list. This solved a lot of the requirements from above
   • Follow-Up: What if each hub had a time-to-expire (TTE)? When it's time, the hub needs to be removed automatically and the robot is NOT allowed to access it anymore.
   • Overall: Felt like this could be improved a bit. I did get stumped a bit for the follow-up, but my interviewer was very collaborative and suggestive. Also felt like we were solving this challenge together. I felt like this round resembles on the job.
   • If I had to score myself: 3.75/5

4. Pure LC Round

   • Warm-Up question: https://leetcode.com/problems/moving-average-from-data-stream/
   • Follow-Up question: https://leetcode.com/problems/k-closest-points-to-origin/
   • Overall: Felt good on both. I've seen both so I was able to solve both in 35 minutes. However, my interviewer wanted a faster time than the heap for the follow-up question. If I hadn't already seen it, I would've been like WTF?! So, this is definitely pure luck round. I did NOT feel like this round helped me with real-skills on the job.
   • If I had to score myself: 4.5/5

I want to stress that EACH ROUND BEGAN WITH 15 MINUTES OF LP. Please do not overlook this. Typical tell me about a time...

Results: Heard back the next day with a passing with flying colors. I seriously did NOT expect it to be this fast nor this positive for feedback. There was a lot of things that I felt like I could've done better or shouldn't have done to begin with.

LC Count: ~130 easy, ~140 medium, ~20 hard
* This is over the span of a year and a half
* I was also prepping for G and Meta
* It was very very very hard...I still can't do DP correctly lol :/

Will update with a link to offer compensation once I get the finalized numbers. Wishing you all the best of luck on your journey!