Unlocking the Secrets Behind Amazon's 60-Second Kindle Delivery: A Deep Dive into System Design

Last question block and and now I'm gonna force the issue cause this was my concern, given your background and the ability to evaluate technical chops as it would pertain to a job at the Target company you're looking at. I'm gonna force the issue by asking you a very specific question that I rarely use um in in these interviews because I love the question so much that I don't want people knowing about the question. But, here we go. So, uh, my previous experience, uh, my last, I guess 22 jobs ago, I ran the Kindle team for Amazon. I was responsible for all reading services, uh, at Amazon. And the original premise, this was before I got there, the original premise of the Kindle, and it remains to this day, is any book, uh, ever published, delivered wirelessly anywhere in the world in 60 seconds or less. That's the promise, right, upon which the entire Kindle empire was built. So, with that as a promise, right? I, I, I'm assuming you've read on a Kindle device on your phone, in the ink or whatever, so you understand that you know, you're in a store and you're going through and you click, I want this book, you click buy and, you know, magic happens, and within 60 seconds, that book is now available for you to read on your device, OK. From a code perspective, and when I say code perspective, I want you to think about how all this has to work, right? From the time I click by. It's the time I see the book on my screen and I can read it. I want you to walk me through everything that needs to happen. To make that a reality. And I will, I will poke and ask kind of clarifying questions, but, but this is a discussion of a large system, and I'm not asking you to design it the way Amazon designed it. I'm asking you to walk me through what you believe, all of the things that need to happen or that you would do to solve that problem.

you are at a very high level, which is normal. Uh, you know, this is, this question is, this question is intentionally vague to understand how you approach the problem.

OK, um. We'll cover that in the, in the follow-up. I'll just move on to the, the, the follow-up question, which is, OK, so latency in the system obviously is a concern, right? So that, that is an issue, right, with this kind of preordained promise of 60 seconds, who knew if that was possible or not, right? But it's, it's OK. So, how would you go about walk as an engineer, right? Your boss comes to you and says, hey, you're the engineer working on this. Walk me through how you would diagnose. The nature of and location of the bottlenecks. In the system in this end to end delivery.

Don't, don't worry about that. You're not, that's, that's not the point of this question. So this is, this is more of a systems design question than it is a system selection question, right?

So if you think about. The what does your, I guess, without any prior knowledge of how the system works, right? Where would you direct your attention? As a first pass. Easy win or likely easy win for immediately speeding up the process. To, to achieve a sixty-second result. Like without saying about where it was falling in the 60-second band, but you just, you know, you, there's a, where would you put your attention to identify likely largest gains that you could apply. To speed up the process

You would focus on the things that were taking the longest, or you would only focus on the things that were taking the longest in the failure cases.

what if with that as a premise,

your failures were being caused.

But you're defining failure as what here, right? Because, because if you've defined a pipeline of events that need to occur in, in some order, I'm not, I'm not applying a cardinality to it or or ordinality to it, I'm just saying some order. That failure is 60 seconds plus. What is failure for any individual component in that pipeline if the the steps are generally always the same. Who do you assign the blame to?

So, I, I guess I'm, I'm gonna press for like a specific like just thinking about what would need to occur for all of this to happen, right? You've talked about off, you talked about credit card, you talked about transmission over protocol, you talked about a bunch of things, right? Mhm. Where like a manager is going to come to you and say, what, OK, great, but what are you gonna go look at, right? It's like in, in a company, they're gonna say, that all sounds great theory, but like, what is your gut telling you? Where are you gonna go spend your time? Or, or how can you think creatively to remove steps that don't need to be there at that time.

So what if I told you that credit card processing as an example, always took 6 seconds, right? So if you think about the, the, the band diagram of of time consumption in the 60 seconds, right? And, and you're assuming that there's, there's kind of a The 60 seconds, there's, there's, it's, you know, you have to think of it as a Gantt chart in some ways, right, because there's some things that can happen in parallel and some things that can't, but that's fine. But removing that and let's just say your total time. Consumed this was a a a consistent big ticket item was credit card processing. What could you do about that?

OK.