Uber Interview Process & Rounds
In candidate reports, Uber's software engineering process is a recruiter call, an online assessment on HackerRank for many candidates, a one-hour technical phone screen, and a virtual onsite of four rounds: two coding rounds, a system design round, and a hiring-manager round that pairs behavioral questions with a deep dive on a past project. Reports that give dates run from under three weeks to about two and a half months end to end, and from June 2026 many candidates describe slow team matching, filled positions or a hiring freeze after passing. Where candidates fell short, it was mostly system design and the hiring-manager round: several who solved every coding problem were still rejected or offered a lower level.
Part of the full catalog of Uber interview questions, covering every round.
AceOffer is not affiliated with, endorsed by, or sponsored by Uber. Questions are reconstructed from interview experiences candidates shared, not from Uber materials.
Key facts
- •7 distinct round types
- •83 questions reconstructed from 302 candidate reports
- •Reports span Oct 2025 – Oct 2026
- •Refreshed monthly · last updated October 2026
The Uber loop, from candidate reports
Reported Uber software engineering loops are mostly algorithm and object-oriented coding rounds plus system design. Some candidates start with an online assessment of algorithm problems; the most reported are minimum edge reversals on a tree and a jump game with prime steps ending in 3. The phone screen is a coding round, often with time on a past project, and the problem is most often the first unique number in a stream, robots matching distances to the nearest blockers, or the longest subarray within an absolute difference limit. Onsite coding mixes LeetCode-style problems, such as Number of Islands with lands added over time and Meeting Rooms II, with object-oriented design such as a parking lot and a meeting scheduler over a fixed list of rooms. System design rounds often center on Uber and Uber Eats products: the Uber Eats shopping cart, a driver heatmap, nearby restaurants and ride hailing. Many loops include a hiring-manager round of about 75 minutes that pairs behavioral questions with a project deep dive, where a diagram is expected and some candidates bring slides. ML Engineer loops add ML coding rounds, such as linear and logistic regression from scratch.
Uber interview timeline
| Step | Typical timing |
|---|---|
| First recruiter contact to phone screen | About two to seven weeks (4 reports), including the time to find a matching team or posting |
| OA to result | Two days to a week (5 reports); one rejection came 15 minutes after submitting, another three weeks later |
| Phone screen to result | Next day to about a week (9 reports); some candidates waited a month or never heard back |
| Phone screen to onsite | Two to about five weeks (5 reports), mostly waiting for interview slots |
| Onsite to decision | Two days to a week (7 reports); others waited two weeks or more |
| End to end | Under three weeks to about two and a half months (4 reports) |
The fastest report went from recruiter to offer in under three weeks, with the offer the Monday after a Thursday onsite. The slow cases are mostly scheduling and silence: candidates describe recruiters taking two weeks or more at each step, onsite rounds spread over a week because an interviewer could not be found, and rounds booked outside the times they had offered. Some candidates were told after a good phone screen that the position had been filled and were moved to another one. Many candidates complain about recruiter speed in particular, and one found their recruiter's email address no longer worked partway through the process.
The Uber interview process, step by step
1. Recruiter call
A short recruiter call in most reports; some candidates got the OA link straight after applying, with no call.
In one report the recruiter asked about years of experience, immigration status, skills, location preference and pay expectations, and said an OA would follow. Uber hires by team or org: one recruiter said a whole org had fewer than six openings at the candidate's level, and another had the candidate find a matching posting and apply to it after the call. Experienced candidates are often placed a level below their title: two candidates with seven years of experience interviewed for SDE2, and a senior engineer with over ten years could only interview for L5a. Invitation emails name the round types: one listed a 75-minute past-experience round and asked the candidate to prepare multithreading, and others described a screen as coding plus design and architecture.
- →Ask early which level you are being considered for. Several candidates learned only after the onsite that they would be offered a lower level.
- →Read the scheduling emails closely, since they name what each round covers. Ask about any round you do not recognize; one candidate asked how to prepare for the past-experience round and got no answer.
- →Check the times the coordinator sends back. Two candidates were booked outside the windows they had given.
2. Online assessment
HackerRank. For experienced roles, two problems in 90 minutes with camera, full-screen sharing and a single-monitor check; new grad versions have three problems in 120 minutes without camera.
Twelve reports describe two problems in 90 minutes, and candidates describe one as medium and one as hard. The most reported problems are minimum edge reversals on a tree (13 reports) and a jump game with prime steps (11 reports), followed by a palindrome paths tree problem, a pipeline throughput problem and a ball-absorbing problem. Several candidates say both of their problems had already been posted on the forum. Each problem has hidden tests, 15 per problem in one report, and some tests show only pass or fail. The passing bar is unclear: candidates moved on after passing 10 of 15 or about 70% of the tests on the second problem, while another was rejected 15 minutes after submitting with a few large cases failing. One late-2025 report describes an older CodeSignal assessment of four questions instead. One candidate had no OA at all and went straight to the phone screen.
- →Write tree and graph traversals iteratively. Three candidates hit recursion depth limits with recursive DFS on the tree problems.
- →Aim for the optimal complexity before you submit. Several candidates passed the visible tests but timed out on large hidden cases with a slower approach.
- →Check the indexing and the expected output in the statement. Two candidates note that an input was 1-indexed.
- →Use one screen. Two reports say only one monitor was allowed, and a reply in another says HackerRank detects extra monitors in the OA and the phone screen.
3. Technical phone screen
About 60 minutes on HackerRank with an engineer: often 15 to 30 minutes on your background or a project, then one coding problem. Some invitations say 45 minutes of coding and 15 of design and architecture.
The coding is usually one problem with follow-ups, though some candidates, senior ones among them, got two problems in the hour. Among the most reported screen problems are first unique number in a stream, a robot grid question and number of islands. Some screens are object-oriented design instead, most often a parking lot. In most reports candidates write and run their own test cases; in one, a dry run of the given example was enough. Follow-ups push toward production: millions of records, a million inputs a second, locks for concurrent queries, and when to clear stale data. Several statements were deliberately thin, sometimes only a few examples, so candidates had to ask for the rules. Solving the problem was not always enough: at least five candidates who finished, two of them bug-free in about 15 minutes, were rejected, and one recruiter said the reason was that the candidate never raised failure cases or edge cases on their own.
- →Write your own tests as you go. Most screens expect runnable code checked against cases you write.
- →Have one project ready for 20 to 30 minutes of questions. Candidates describe detailed, challenging questions on their project in the screen, and one says it was harder than they expected.
- →Bring up scale, concurrency and failure cases before you are asked. The follow-ups go there anyway, and one rejection was explained by their absence.
- →Clarify the inputs and outputs before you code. One candidate who started before understanding the problem ran out of time, and another spent 20 minutes pinning down a thin statement.
4. Onsite: coding (two rounds)
Two rounds on HackerRank, held virtually in recent reports, some opening with time on your resume.
Most onsite reports describe four rounds, two of them coding. One coding round is often a LeetCode-style problem such as number of islands, alien dictionary or LRU cache, and the other is often object-oriented: a parking lot, a meeting scheduler over a fixed set of rooms, or a pricing engine. Interviewers move quickly to follow-ups: one candidate says the main problem took under half of each 45-minute round and the rest was added requirements. Follow-ups lean toward production, such as a million requests a second, thread safety and very large input. Some rounds have two problems. Interviewers judge code quality as well as correctness: one kept asking a candidate to remove variables and simplify a working solution, and another expected each part solved before seeing the next. In 2026, two reports describe an AI-assisted coding round: one on HackerRank with a built-in assistant, finding a money-transfer bug in a running app with a front end and back end, and one an object-oriented design round with AI. Other 2026 candidates say AI was not allowed in any round.
- →Practise union-find until you can write it from memory. Several candidates lost time on it because they had not written it recently.
- →Solve the part in front of you before designing for later parts. One candidate says an interviewer did not accept a design aimed at the last function before the earlier ones were done.
- →Prepare a parking lot and a meeting scheduler as runnable classes, with tests. They are the most reported object-oriented problems.
- →If your loop has an AI-assisted round, ask which languages it supports. One recruiter said only Java and Go.
5. Onsite: system design
One round; in some loops run by a staff engineer or by the bar raiser.
The most reported prompts are Uber products: the Uber Eats shopping cart (7 reports) and a driver heatmap (7 reports), then a chat system like Messenger (4), nearby restaurants for Uber Eats (3), ride hailing (3) and stock price alerts (3). Other prompts in reports include Uber Eats search, popular items, a URL shortener, a payout system and a distributed file system. Prompts often start vague, and interviewers watch whether you drive the requirements yourself. Deep dives in reports cover why Kafka, stream versus batch processing and how to avoid losing data, how to store and query historical location data, concurrent cart updates from several devices, and neighboring cells in geospatial lookups. Two rejections were about the conversation rather than the diagram: one candidate was told to talk with the interviewer more instead of only giving answers, and another felt the interviewer wanted one particular design.
- →Prepare the Uber Eats cart and the driver heatmap first; together they account for 14 reports.
- →Drive the session: settle scope, scale and the main read and write paths before you draw, and check in with the interviewer as you go.
- →Know the internals of what you propose. One candidate had to defend why Redis could handle the throughput, and another discussed the Kafka partition key and wrote stream-processing pseudocode.
- →Prepare for an open-ended data platform prompt too. A staff interviewer asked one candidate to design a system for analyzing app user behavior with almost no stated requirements.
6. Onsite: hiring manager (behavioral and project deep dive)
Usually with the hiring manager; 75 minutes in the reports that give a length, with behavioral questions woven into a deep dive on a past project.
Candidates walk through a past project, usually with a high-level diagram: its scope, the hard parts, the trade-offs and what they would change. One candidate learned only in the round that a diagram was expected and drew it on the spot. Behavioral questions in reports include conflict, mentoring juniors, a failure, ownership, feedback you disagreed with, why you are leaving and why Uber. A manager candidate was asked about managing a poor performer and about disagreements with a manager and with other functions. This round carries weight: one candidate's recruiter said the hiring manager vouched for them and pushed back on the bar raiser after an average system design round, and they got an offer, while other candidates who did well technically were rejected for weak leadership or collaboration signals, or told they were below the level's bar.
- →Prepare one project you can discuss for about 25 minutes, with a diagram ready. A candidate who got an offer recommends slides and a project that fills that time.
- →Prepare mentoring and leadership stories even for mid-level roles. One candidate skipped mentoring although Uber's preparation guide lists it, and the feedback said leadership was not strong enough.
- →Be ready to defend the timeline and scope of your work. One interviewer was unhappy that a complex API took two weeks, and a manager challenged another candidate's project.
- →Pick a project close to the team's area. One ML candidate was told their GenAI project did not match a pricing team.
How the Uber loop differs by role
New grad (SDE I)
New grad reports describe a three-problem HackerRank OA in 120 minutes with full-screen sharing and no camera, which one candidate thinks draws from a random pool. One candidate had a phone interview between the OA and the final rounds, mostly coding with their own test cases. Another had 45 minutes of coding and 10 minutes of behavioral questions, and was rejected after completing and extending the problem. One candidate passed the OA two days after taking it and then waited more than a week for an interview slot.
Interns
A 2027 intern OA had three problems. An Amsterdam intern had an OA of medium problems and then a final round of two coding interviews, one on currency conversion rates and one a simple DFS. A PhD ML intern had an OA they call mostly easy, one coding interview on meeting rooms, one ML interview on optimizing a quadratic function, and then team matching.
ML engineer
ML engineer onsites in reports have four rounds: an ML system design (Uber Eats restaurant recommendation, feed ranking or fraud detection), an ML coding round (linear and logistic regression from scratch, in one report required to run), an algorithm coding round, and a hiring-manager round. Phone screens vary by team: a convex function minimization for a marketplace pricing team, multi-head self-attention in PyTorch for an AI team, a bigram text generator, and coding plus ML knowledge questions. One candidate notes that each team seems to ask different questions, and two say the project discussion took up much of the screen. One L5a candidate was told they might be offered L4 and stopped the process.
Staff
One staff candidate had two rounds, Scope & Impact and Architecture, both built on their past projects, and was told staff engineers must rethink problems at the root rather than extend existing solutions; they did not pass. Another staff onsite had five rounds, including an AI-assisted coding round and a technical communication round, and ended in a rejection without a lower-level offer. One staff phone screen opened with a behavioral question about changing a system design for a trade-off before the coding.
Frontend
One frontend loop had a phone screen simulating a tournament from player ranks, then an onsite with a calendar design, behavioral questions with detailed UI questions, an algorithm round and a parking lot built in React. Another frontend candidate got a graph problem in the phone screen.
Backfill roles
One L5a infrastructure candidate in a backfill role had a shorter loop: a phone screen and then three onsite rounds, coding, system design and the hiring manager, who said the candidate was already a top candidate from the earlier feedback.
Scientist roles
Scientist and applied science reports describe rounds built around case studies and experiment design, with light coding, which this guide does not cover.
After the final round
Results come from the recruiter, two days to a week after the onsite in most reports that give timing; some recruiters gave round-by-round feedback, while others said they could not share any. Passing does not guarantee an offer. Uber hires by team, and candidates describe the original team choosing someone else and a match with another team about a month later, roles filled before or after the onsite, and team matching that dragged on for one to three months. In two threads, candidates interviewing for L5a passed but were offered L4, and replies in both describe the same outcome. From June 2026, many candidates mention a hiring freeze or pending headcount; one candidate's process paused for about a month after the phone screen and then resumed. After a failed onsite, one candidate mentions a six-month cooldown, and after a failed OA one candidate was picked up by a different team.
What does Uber ask in each round?
Uber interviews span 7distinct round types, shown below. Counts reflect distinct questions per round across the loops we’ve indexed.
- •Knowing Union-Find well. Two reports say interviewers expected it on the islands questions, even where DFS would work.
- •Planning for the scale follow-up. First Unique Number and the hit counter are followed by questions on very large input, a million writes per second, locks and when to clear stale data.
- •Moving past brute force. The robot blocker screen is followed by precomputing distances in all four directions, and in one onsite the longest-subarray problem was pushed from brute force to O(n log n) and then O(n).
- •Writing runnable object-oriented code. One parking lot report asks for fully runnable code, with data structure choices you can defend.
- •Recursive DFS on the tree OAs. Reports on the edge reversal and palindrome paths problems say the recursion depth fails, so traverse iteratively.
- •Brute force on the OA. On the ball-absorbing problem, candidates who did not group by row and by column timed out on hidden cases.
- •Coding before the statement is pinned down. Several OA problems are reported under different names and wordings, so check direction, indexing and the expected output first.
- •Assuming the LeetCode version. The reported Edge Length Limited Paths condition differs from the LeetCode one, and the squares problem is reported both as kth largest and kth smallest.
How to prepare for the Uber interview
1. Practise HackerRank tree and graph problems on a 90-minute clock
Work medium and hard tree, graph and binary-search problems with large hidden tests in mind, write traversals iteratively, and read each statement for indexing and output format before coding.
2. Drill the most reported coding problems, then their follow-ups
Number of islands with union-find, first unique number, the robot grid question, meeting rooms and LRU cache come up again and again. For each, practise the scale, concurrency and production follow-ups out loud, not just the first solution.
3. Write object-oriented designs as runnable code
Build a parking lot and a meeting scheduler as classes with your own tests on a 45-minute clock, then add a requirement and see how much you have to change.
4. Prepare Uber product designs
Cover the Uber Eats cart, a driver heatmap, a chat system, nearby restaurants and ride hailing, with geospatial indexing, stream processing and data loss ready for the deep dive. Practise leading a vague prompt from requirements to a design.
5. Build a 25-minute project story with a diagram
Pick a project close to the team's work, draw its architecture, and prepare its scope, trade-offs and your decisions, plus stories on conflict, mentoring, failure and feedback.
FAQ
How long does the Uber interview process take?▾
Is there an online assessment?▾
How many rounds are in the Uber onsite?▾
What is the 75-minute past experience round?▾
Can I use AI in the Uber interview?▾
Does Uber down-level candidates?▾
Is Uber hiring, or is there a hiring freeze?▾
Does Uber have a question bank?▾
How fresh is this Uber interview data?▾
Get the full Uber catalog
Every question, every candidate-reported follow-up, and what passers actually do. Monthly refresh.