Autocoder

Uber Interview Questions and Answers for SDE Roles

Last updated 2026-11-19

Uber's interviews emphasise practical coding and, for experienced roles, design problems drawn from marketplace systems such as matching, location and payments. This page works through that mix with model answers.

The questions are practice in the style of topics Uber describes, not a record of past interviews.

Coding style, per UberPractical problems, working code
Format describedLive collaborative interview, then a team interview
Design themesLocation, matching, payments, queues
RolesSDE I, II, Senior, intern programme

How the rounds are described

Uber's How we hire page says technical roles include a live collaborative problem-solving interview, some roles a job-related exercise, then a team interview. Uber's engineering blog, which is about five years old, says its coding interviews run about an hour with 30 to 40 minutes on a practical coding exercise, using practical problems rather than puzzles and looking at how you turn thinking into working code. A candidate-reported backend loop included three coding interviews, two design or architecture interviews and one collaboration or leadership interview; that account is anecdotal.

Uber lists roles from Software Engineer I and II through Senior Software Engineer, and a Software Engineering Internship Program. Round counts for your role may differ, so confirm with your recruiter.

Related practice

Coding problems in the style Uber describes are in Uber coding round questions. For the process see how to crack the Uber interview.

Hiring status

In May 2026 Uber was reported by People Matters to be planning new Bengaluru (about 5,000 seats) and Hyderabad (about 4,600 seats) campuses totalling nearly 9,600 employees, opening by the end of 2027; the article gives no hiring numbers by role. Uber's emerging-talent pages list a Software Engineering Internship Program and Data Science internships, including Bangalore. See Uber hiring for the sourced summary, and apply through Uber's own careers site, uber.com/in/en/careers.

Uber Interview Questions and Answers for SDE Roles: questions with worked answers

  1. 1

    How would you find the nearest available drivers to a rider?

    hard

    Index driver locations in a spatial structure instead of scanning all drivers. Common choices are a geohash or a grid cell id (S2 or H3 style hexagons) mapped to the set of drivers in the cell, a quadtree that subdivides dense areas, or a k-d tree for static points. Query the rider's cell and its neighbours, filter to available drivers, then rank by estimated travel time and not just straight-line distance. Drivers move constantly, so updates must be cheap, which favours cell-based indexes in memory with periodic location pings.

  2. 2

    Sketch the design of a ride-matching service.

    hard

    Clarify scale and goals: low latency, no double assignment. Riders request a trip, a matching service queries nearby available drivers, ranks them (ETA, rating, acceptance chance), offers the trip to one driver (or a few in sequence) with a timeout, and on acceptance marks both parties as matched atomically. Location updates flow through a streaming pipeline into the spatial index. Discuss partitioning by city or region, what happens if a driver does not respond, and how to keep state consistent if a service crashes mid-match. See system design interview.

  3. 3

    How do you prevent two riders from being assigned the same driver?

    hard

    Make the assignment an atomic compare-and-set on the driver's state: change status from available to reserved only if it is currently available, in a single operation in a store that supports it (a conditional update in a database, a Lua script or compare-and-set in a key-value store, or a lock keyed by driver id with a short TTL). Whoever loses the race gets a failure and tries the next candidate. Add a timeout on reservations so a crashed matcher does not hold a driver forever.

    python
    import threading
    class Driver:
        def __init__(self):
            self.status = 'available'
            self._lock = threading.Lock()
        def try_reserve(self):
            with self._lock:
                if self.status != 'available':
                    return False
                self.status = 'reserved'
                return True
    java
    import java.util.concurrent.atomic.AtomicReference;
    class Driver {
        private final AtomicReference<String> status = new AtomicReference<>("available");
        boolean tryReserve() { return status.compareAndSet("available", "reserved"); }
    }
  4. 4

    How would you estimate the time of arrival between two points on a road network?

    hard

    Model roads as a weighted graph where edge weights are travel times, and run a shortest-path algorithm: Dijkstra, or A* with a straight-line heuristic to explore fewer nodes. At city scale, plain Dijkstra is too slow per request, so systems precompute structure (contraction hierarchies or partitioned graphs) and adjust edge weights with live traffic. Mention that real ETA also uses learned models on top of the route time, and that fallbacks are needed when data is missing.

  5. 5

    What is a reasonable approach to surge pricing?

    medium

    Divide the city into cells and compare demand (open requests) with supply (available drivers) per cell over a short window, then map the ratio to a price multiplier that increases when demand outstrips supply, with smoothing and caps so prices do not jump wildly. Show the multiplier before the rider confirms. Mention data freshness, fairness and regulation, and testing changes with experiments before wide rollout.

  6. 6

    Exactly-once versus at-least-once delivery: how do you handle payments?

    hard

    True exactly-once delivery across a network is not achievable; systems provide at-least-once delivery and make processing idempotent so duplicates have no effect. For payments, attach a unique id to each charge, store processed ids in the same transaction as the ledger entry, and ignore repeats. Use a transactional outbox so a database change and its event publish cannot diverge. Reconcile with the payment provider periodically to catch gaps.

  7. 7

    How do message queue partitions affect ordering?

    medium

    In a partitioned log such as Kafka, ordering is guaranteed only within a partition. If all events for one trip must be processed in order, key messages by trip id so they land in the same partition, and process each partition with one consumer at a time. More partitions give more parallelism but no global order. A consumer group rebalance can cause duplicates, so handlers should be idempotent.

  8. 8

    Request latency for a service has jumped. How do you investigate?

    medium

    Start with the scope: which endpoints, regions and clients. Check recent deploys and config changes, then dashboards for the usual suspects: CPU, memory and garbage collection, thread-pool or connection-pool saturation, slow dependencies and database queries, and cache hit rate. Use distributed traces to find the slow span. Mitigate first by rolling back or shedding load, then fix the cause and add an alert or test that would have caught it.

  9. 9

    Uber's engineering blog says its coding interviews use practical problems. How do you show that in an interview?

    medium

    Uber's engineering interview post describes coding interviews of about an hour, with 30 to 40 minutes on a practical coding exercise, where interviewers look at how you turn thinking into working code. Show that by clarifying the requirements, choosing simple data structures, writing readable code that runs, testing it with an example and then discussing improvements. Avoid jumping into an elaborate design before something works.

  10. 10

    Tell me about a time you took ownership in an ambiguous situation.

    medium

    Describe a situation without clear owners or requirements, how you decided what to do first, who you aligned with, what you delivered and how you checked it worked. A candidate-reported Uber backend loop included a collaboration and leadership interview, so be ready to show how you worked with others, not only individually. Use the structure in the behavioural interview guide.

  11. 11

    Tell me about a disagreement with a teammate over a technical decision.

    medium

    Explain both positions fairly, how you gathered evidence (a benchmark, a prototype or user data), how you reached a decision and how you supported it afterwards. Avoid framing it as winning or losing; interviewers want to see that you can disagree productively and then commit.

  12. 12

    Describe a bug that was hard to find and how you eventually found it.

    easy

    Walk through the symptom, your hypotheses, how you narrowed them down (logs, minimal reproduction, bisecting changes) and the root cause. State the fix and what you added to prevent a recurrence, such as a test or alert. A methodical process is the point; do not just say you stared at the code.

Frequently asked questions

Are Uber coding interviews LeetCode style?

Uber's engineering post describes practical coding problems rather than puzzles, with emphasis on working code. Data structure fluency still helps.

Does Uber hire software engineers in India?

Uber was reported in May 2026 to plan new Bengaluru and Hyderabad campuses, and lists internships including Bangalore. See the sourced summary under Uber hiring.

Is there a system design round at Uber?

Experienced-role accounts describe design or architecture interviews. Early-career loops may focus more on coding.

Where can I find Uber's current openings?

See Uber hiring and uber.com/in/en/careers.

Sources

Related guides

More in Company Interview Questions

See all in this section