Uber Interview Questions

You have found a question list and still cannot tell which entries matter for the role you applied to. This section sorts Uber's questions into the buckets its rounds actually use, so you know what to rehearse first. There are three for engineering roles, and non-engineering loops swap the first two.

Uber hires across ride-hailing, delivery, freight, and internal platform teams, and the loop reflects that spread:

• Coding: data structures and algorithms in a shared editor, with a pull toward graphs, heaps, sliding windows, and interval merging. Routing and location problems appear because they map onto the business.
• System design: distributed systems that change every second. Driver matching, a surge pricing pipeline, or a geospatial index are all common framings.
• Behavioral: ownership, disagreement, and shipping without complete information, usually from a hiring manager or a cross-functional partner.

Analytics cases replace the technical buckets for data roles, product sense and metrics for product management, and operations scenarios for city teams. The question bank hub collects those role-specific lists.

These are the shapes that recur, written the way you would hear them:

• Find the shortest path between two points on a city grid with some roads closed.
• Merge a set of overlapping time intervals representing driver availability.
• Given a stream of trip events, return the top k pickup locations in the last hour.
• Implement an LRU cache, then explain what breaks when it is shared across processes.
• Design a service that matches riders to nearby drivers in near real time.
• Design surge pricing: how the signal is computed, where it is stored, how stale it is allowed to be.
• Design a system that tracks millions of moving vehicles and answers proximity queries.

The design prompts are not asking for a finished architecture. They are asking whether you can name the constraint that makes the problem hard, which for most Uber-shaped problems is that the data is geospatial and changing constantly. Name that early and the rest of the conversation has somewhere to go.

Reading a question list builds recognition, not recall. The gap shows up the moment another person is waiting for you to speak.

A concrete case: a backend engineer preparing for a senior role at a ride-hailing company had drilled graph problems on paper for weeks and still stalled halfway through the driver-matching prompt. The problem was not knowledge. They had never said the words consistent hashing in a sentence with someone listening. Two spoken run-throughs fixed what twenty more silent ones would not have.

A sequence that works:

• Pick one question from each bucket above.
• Answer it out loud, on a timer, with no notes.
• Record it or run it against an AI interviewer, then listen for where you went quiet.
• Rewrite only the part that failed, and run it again.

The SubcueAI mock interview runs that loop with spoken questions and spoken answers, which is closer to the real round than a flashcard app.

Uber runs coding rounds in shared editors and design rounds as conversations, and those two formats answer this question very differently.

In a conversational round on Zoom, Google Meet, or Microsoft Teams, a live assistant does something narrow: it transcribes the interviewer and puts a structure on your screen while you are still forming the sentence. SubcueAI does that from a native macOS or Windows app with a local floating overlay, and from a browser extension side panel that captures the meeting tab's audio. No bot joins the call, and nothing is injected into the meeting page.

The limits matter more than the capability:

• If you are sharing your screen, the overlay is on that screen unless you moved it off the shared area first.
• If the round is proctored or runs inside a locked-down assessment tool, no real-time desktop assistant is appropriate. The security page is direct about which contexts are out of scope.
• A coding round where you type in the interviewer's editor is a round you have to know how to solve.

Which is the useful framing: the assistant carries recall in conversation, and practice carries everything else.