Amazon New Grad Software Engineer Interview Experience — Leadership Principle BQs and an Edit Distance Follow-up

Round 1

BQ1: Give an example of how you organized a product.
BQ2: Give an example of a time you missed a deadline.
BQ3: Talk about a time you thought something was wrong but stuck to your own view.

Prepare your stories using the STAR method. Except for the first two — "the project you're most proud of" type stories — try to add lessons learned to the rest. Tell the story in a concise, easy-to-follow way. The interviewer doesn't really care whether you used Java or Python, or whether you used some advanced piece of tech. What they care about is your attitude in the story, what actions you actually took, and whether the result matched what you expected. If you don't know how to structure your story, you can ask GPT for help.

If you're worried about not answering the BQs well, prepare more examples ahead of time. This works well if you're not very confident speaking on the spot and are afraid of blanking. Take Amazon as an example — prepare six stories, then map each one to a Leadership Principle you can use it for. Conflict obviously maps to Disagree and Commit; time management can map to Bias for Action; a project you're proud of can map to several — Dive Deep, Ownership, Learn and Be Curious.

Round 2

The interviewer was an SDE, felt pretty easygoing overall, and the conversation flowed well. We started by briefly going over the projects on my resume — didn't feel much pressure. He talked pretty fast and would occasionally slip in a filler phrase or two in Chinese, so the whole conversation felt pretty natural.

The coding question was the classic Edit Distance problem. For the basic version you need to write actual code — mainly dynamic programming, defining dp[i][j] as the minimum number of operations needed to turn the first i characters of word1 into the first j characters of word2, then doing the state transitions based on the three operations: insert, delete, replace.

There was a follow-up after that: on top of the original operations, you're additionally allowed to do one reverse of the entire string — just walk through your approach, no need to write code. My take was that it's still overall dynamic programming, just adding in the possible results after a reverse on top of the original recurrence. You can work out the edit distance to the target string separately for the normal string and for the reversed string, then take the minimum; if the reverse itself counts as one operation, you'd need to add that operation's cost to the result. The overall approach is close to the original problem — the key part is folding the new operation into the state or into the set of candidate transitions.