Senior Software Engineer, Apps Interview Experience
The weirdest part for me was the AI coding round. They gave me a buggy maze class in HackerRank, made me use the built-in assistant, and then spent the back half probing why I trusted one AI-generated approach over the others.
Interview process
I interviewed for a Senior Software Engineer role on the apps side, and the process was a short recruiter call, one standard coding screen, and then a five-round onsite. The biggest thing that stood out was that every interview had two people in it, usually one leading and one shadowing, so it felt like they were calibrating in real time. The onsite mixed normal coding and backend system design with an AI coding round in HackerRank where I had to use their built-in assistant to debug and build out a maze problem. The technical communication round was much more casual than I expected and ended after about 30 minutes because the interviewer said he had what he needed. I finished the full loop but didn't get an offer.
- Recruiter screen
- Phone interview
- Final round
Interview tips
I'd spend most of my prep on coding because that seemed to matter a lot, and the questions felt pretty aligned with the usual LinkedIn-tagged LeetCode stuff. For the AI round, don't just let the tool write code and move on. Be ready to explain why you accepted one output over another, how you'd test it, and what you'd change if it had to go into production. For the manager round, I'd have my career story ready end to end, not just STAR stories, and I'd make sure my examples show ownership and line up with LinkedIn's values.
Company culture
They seem to be experimenting with newer interview formats, especially the AI coding round, but the evaluation still feels a little uneven depending on who you get. Every room had two interviewers, which made it feel like they care a lot about calibration. The process was team-independent within apps, and team matching seemed like it would only happen after passing the onsite, so it felt more centralized than team-specific companies. Coding looked heavily prioritized, while the manager round was digging for career progression, ownership, and decision-making. The technical communication round felt like they wanted to see whether I could talk through real work naturally with other engineers, not just recite polished stories.
Questions asked
Overview
My onsite was five rounds with two interviewers in each room, usually one leading and one shadowing, and it mixed normal coding and system design with a newer AI coding round, a manager behavioral, and a surprisingly casual technical communication conversation.
Question types asked
Specific questions asked
Use the AI assistant to debug and complete this maze implementation.
How would you make sure the maze is well formed, including the outer wall?
Given a start and end position, how would you generate the maze so there is always a path?
Why did you choose this AI-generated approach over the other options?
How would you test each change?
If this went to production, what would you change?
They gave me starter code in HackerRank with basically a maze class and a print function, but the maze wasn't really implemented. I used the built-in AI assistant to debug it, make sure the outer wall existed, and then generate a maze with a guaranteed path from start to finish. The AI suggested a few approaches, and I picked a simpler tree-traversal-style one because it was easier to reason about. They kept probing on why I chose it, how I'd build tests, and what I'd change for production, like service boundaries, concurrency, and whether parts could be split up.
How did your system handle tokens correctly?
How was pricing thought about, per query or per token?
What challenges came up that were similar to what we're seeing?
I talked about an AI integration project where I was building a framework that let customers talk to external LLMs. That turned into a long conversation about token handling and how usage gets assigned correctly, plus how pricing might be set, like per query versus per token. It felt much more like an actual working conversation than a formal interview, and both interviewers jumped in with their own experiences.
Walk me through your career progression.
How did you move from intern to full-time?
Did you stay on the same team or switch teams?
Why did you make those moves?
How did your promotions happen?
The manager really wanted my whole career story, not just a few rehearsed STAR answers. I walked through how I started as an intern, converted to full-time, how my scope changed over the years, whether I had switched internally, and why I made those decisions. They also asked specifically about promotions and what I had done to get there, which felt like they were looking for ownership and growth signals.
I answered with a project I was proud of and focused on what I had actually owned and driven. The feel of that round was less about one perfect story and more about whether my choices and progression showed initiative, ownership, and good judgment.
They used this more as an influence and conflict question than a pure conflict question. I framed it around how I would get people aligned on a solution, and it felt like they were checking whether I could persuade peers instead of just pushing my own idea.
What if some teams don't want to put in the effort to move?
What if only one team is left on the old version?
Would you keep the old version running for that team?
I walked through how I would coordinate the migration and get teams to make the required changes, and then they really pushed on the holdout cases. The follow-ups were the interesting part: what to do if teams won't prioritize the work, and what happens when there's just one team left and you're deciding whether to keep the old version alive for them.
This one was scoped as a backend-only system design. I designed an autosuggest or autocomplete system from the backend side rather than talking about UI details.
This coding round was actually pretty straightforward. It was a string problem where I had to reason about allowed edits and check whether the result could be a palindrome.
Get full access with a membership, or share your experience to try it free.
