amazon sde summer intern 2026 onsite interview experience

First in-person interview I have ever done

It wasn't bad, I thought because of Amazon's hiring spree and invitation to on-site interview, the hiring criteria requirement would be lax, but it wasn't.

Here was the interview process and experience:

Interview process

2, 60-minute interviews back to back  

(45 minutes technical, 15 minutes behavioral/resume grill/AI questioning)

First interview

15 minutes:  

Talking about resume, projects I've done, previous internship experience.

45 minutes...

First in-person interview I have ever done

It wasn't bad, I thought because of Amazon's hiring spree and invitation to on-site interview, the hiring criteria requirement would be lax, but it wasn't.

Here was the interview process and experience:

Interview process

2, 60-minute interviews back to back  

(45 minutes technical, 15 minutes behavioral/resume grill/AI questioning)

First interview

15 minutes:  

Talking about resume, projects I've done, previous internship experience.

45 minutes:  

Similar to system design kind of question, given multiple data structures, must augment the maps to extract certain features, get the elements that fulfill a certain criteria, etc. Wasn't Leetcode style, but wasn't a system design either. Almost like generic coding ability.

Second interview

45 minutes:  

Question similar to encode and decode strings. Given a string, objective is to replace all instances of ":x:" with "y". every key value would start with ";" and end with ":", which he had explicitly stated. Using this, my original solution was an iterative approach, where upon encountering a ";", I checked all possible keys and would see if the actual end of the key I'm checking matched. if it did, then replace and move on. had some issues with the string replacement as I was using a little bit more memory, didn't know the solution to getting optimal space (instead of using strings, make it an array and then use .join). I had an almost optimal solution but was really nervous and overcomplicated it. We walked through the problem, and I felt very nice talking to him and had a very positive experience (thought I did good in that), but apparently not.

15 minutes:  

Asking about typical behavioral questions, talked a little bit about how I use AI to code. Also talked about C/C++ since I just finished a course in that. It was super nice since he also used C/C++ and asked me questions about what tools I use to debug (not AI) — valgrind, GDB, etc.

AI-heavy questions.

Overall, the experience was not bad. It's just that I guess I didn't meet the bar. Very disappointed.