Microsoft Software Engineer Interview Questions

Before designing it, I would clarify the requirements: vehicle types, spot types, entry and exit gates, payment flow, and whether the system needs to support reservations. At a high level, I would model the core entities as parking lot, floor, spot, vehicle, ticket, and payment. I would keep the design flexible so that adding a new spot type or pricing rule does not require rewriting the whole system. The trade-off is between keeping the model simple and supporting future changes, but I would lean toward a clean object-oriented design with clear responsibilities. If scale were important, I would also talk about concurrency for spot allocation and how to avoid double-booking.

Before designing it, I would clarify the channels, delivery methods, latency expectations, and whether notifications must be personalized or only based on events. I would start with an event ingestion layer that receives mentions, replies, or alerts, then fan those events out to a notification service that decides who should receive what. I would store notification state in a database so users can read, dismiss, or mark notifications as seen across devices. The reason I would use asynchronous processing is that notifications should stay fast even when message volume spikes. I would also think about retries, deduplication, and fallback delivery so the system remains reliable under load.

I would first clarify whether the edits need to appear in real time, whether offline editing is required, and how strict the conflict resolution rules should be. At a high level, I would use a document service that stores the source of truth, a collaboration layer for real-time updates, and a sync mechanism for clients connected at the same time. For conflict handling, I would consider operational transform or CRDTs depending on the latency and consistency goals. The reason I separate storage from collaboration is that persistence and live editing solve different problems. I would also mention version history, permission control, and recovery from disconnects because those are important in a real product.

I would first think about using a set to track characters as I scan the string. If I see a character that is already in the set, then I can return false immediately; otherwise, I add it and continue. If the interviewer asks for even less space, I would mention that a bitmask can work when the input alphabet is small and fixed. The reason I choose a set first is that it is simple and gives O(n) time with O(n) space. I would also mention edge cases like an empty string and whether the input is case-sensitive.

I would first clarify whether the limiter should be per user, per app, per endpoint, or per region, because the policy changes the architecture. My default design would use a token bucket model because it handles bursts better than a strict fixed window. I would place enforcement close to the request entry point and store counters in a fast shared system so multiple servers can make consistent decisions. The trade-off is between accuracy and latency, so I would keep the decision path lightweight and make the state store highly available. I would also plan for what happens when the limiter store fails, because that failure mode needs to be defined up front.