Uber SDE-2 (L4) India 2025: 6-Round Gauntlet with Job Scheduler Machine Coding to Offer

TL;DR. 6 rounds. Not 4, not 5, six. CodeSignal screening (cleared in 30 minutes), machine coding with a job scheduler involving concurrency and multithreading, a LeetCode-hard Segment Trees problem-solving round, a system design round on Facebook Messenger with evolving requirements, a hiring manager grilling on project architecture, and a bar raiser who interrogated every design trade-off the candidate had ever made. Result: offer. Compensation undisclosed due to NDA. Source post linked at the bottom.

This is not a "How to crack Uber SDE-2" compilation. This is one specific candidate's 6-round loop at Uber India in 2025, with the actual round structure, the actual machine coding topic, and the actual system design prompt. The candidate signed an NDA, so exact question wording is absent, but the shape of every round is here.

The verified candidate

The NDA means no name, no company, no compensation figures. What remains is the full structure of the loop, which is the more useful signal for preparation. You cannot study someone else's offer letter. You can study the 6 rounds they went through.

Why Uber runs 6 rounds for SDE-2

Most product companies in India run 4 rounds for SDE-2: one DSA, one LLD or machine coding, one system design, one hiring manager. Uber adds two: an independent problem-solving round (separate from the machine coding DSA) and a bar raiser.

The bar raiser at Uber serves the same purpose as Amazon's: a calibration interviewer from a different team with veto power. The additional problem-solving round (Round 3) exists because Uber separates "can you build working software" (machine coding) from "can you solve algorithmic puzzles" (PS). Other companies combine these into one or two rounds. Uber keeps them distinct, which means you need to prepare for both independently.

The total interview time is approximately 6 hours of live rounds, spread across multiple days. Add the CodeSignal screening and the scheduling logistics, and the full pipeline from first screen to offer typically runs 3 to 5 weeks.

Round 1: CodeSignal Screening

The screening is the lowest bar in the loop. LeetCode Easy and Medium problems, no tricks, no advanced data structures. The candidate finished in 30 minutes of a 60-minute window, which means the problems were well within their preparation range.

The purpose of this round is elimination, not calibration. Uber uses CodeSignal to filter out candidates who cannot write working code under time pressure. If you are targeting SDE-2 roles, this round should not consume preparation time. If it does, the rest of the loop will be significantly harder.

One practical note: CodeSignal has its own IDE and submission environment, which differs from LeetCode. If you have never used CodeSignal before, do a practice session on the platform before the real screening. The syntax highlighting, autocomplete, and test-case runner are all slightly different, and those differences cost minutes under time pressure.

Round 2: Machine Coding (the hardest round)

This is the round the candidate specifically called out as the most challenging. Design and implement a job scheduler that handles concurrent job execution with multithreading.

The exact problem statement is under NDA, but the shape is clear: you receive jobs with priorities, dependencies, and execution times. Your scheduler must run jobs concurrently up to a thread pool limit, respect dependency ordering, handle job failures and retries, and report job status. The interviewer adds requirements incrementally during the round.

What "machine coding" means at Uber

Machine coding at Uber is not LLD whiteboard. You write real, running code in a shared editor. The code must compile, execute, and handle edge cases. The interviewer adds new requirements at the 30-minute mark and again at the 45-minute mark, testing whether your initial design can absorb change without a rewrite.

The job scheduler specifically tests three things that separate SDE-2 from SDE-1:

1. Concurrency primitives. Can you use thread pools, locks, semaphores, or concurrent queues correctly? Not theoretically, but in code that runs without race conditions.
2. Incremental design. When the interviewer adds "now jobs can have dependencies" at the 30-minute mark, does your code absorb it or collapse?
3. Edge case instinct. What happens when a job fails? What happens when a dependency cycle exists? What happens when the thread pool is exhausted?

Preparation approach

Practice machine coding problems that involve concurrency. The standard set: thread-safe producer-consumer queue, rate limiter, task scheduler with priorities, in-memory key-value store with TTL. Write each one from scratch in your language of choice. Time yourself to 60 minutes. The critical skill is not knowing the patterns but typing them fast enough under time pressure with evolving requirements.

The candidate passed all edge cases. This means they had practiced concurrency-based machine coding before the interview, not for the first time during it.

Round 3: Problem-Solving (Segment Trees, Hard)

A pure algorithmic problem-solving round, completely separate from the machine coding round. The topic was Segment Trees, which is the upper end of what SDE-2 candidates are expected to handle.

Segment Trees are a data structure for efficient range queries and point updates on arrays. Problems in this category include range sum queries with updates, range minimum/maximum queries, lazy propagation for range updates, and interval-based problems like counting points within a range.

The candidate solved the problem but needed hints from the interviewer. At Uber, needing hints in the PS round is not an automatic failure. The interviewer is assessing whether you can incorporate hints productively (take the hint, pivot, reach the solution) versus whether hints bounce off you (take the hint, still stuck, need another hint). One or two hints leading to a complete solution is a pass. A cascade of hints leading to a partial solution is a fail.

Why Segment Trees at SDE-2?

Segment Trees are rare in SDE-1 interviews because most SDE-1 candidates have not seen them. At SDE-2, the assumption shifts: you have 2 to 5 years of experience, and you should have encountered range query problems in production or competitive programming. If you have not, this round is where the gap shows.

The preparation approach: solve 10 to 15 Segment Tree problems on LeetCode, covering basic range sum, lazy propagation, coordinate compression, and 2D segment trees. If you can solve a lazy propagation problem in under 30 minutes, you are prepared for Uber's PS round.

Round 4: System Design (Facebook Messenger)

Design a chat application similar to Facebook Messenger. The initial requirements include one-on-one messaging, group chats, message delivery status (sent, delivered, read), and online/offline presence indicators.

This is a classic system design prompt, but Uber's version adds evolving requirements. Midway through the round, the interviewer added new constraints (the specific additions are NDA-protected, but the pattern is consistent with Uber's approach: "now add message search across all conversations" or "now handle 10 million concurrent users" or "now add end-to-end encryption").

The key design components

The candidate passed but had to modify their schema design during the round as new requirements arrived. This is the expected outcome. Nobody designs a perfect Messenger system in the first 15 minutes. The interviewer is testing whether you can evolve your design under constraint changes without throwing away your earlier work.

Preparation for evolving-requirements system design

Uber's system design rounds consistently add requirements mid-round. Prepare for this by practicing system design with a partner who interrupts you at the 20-minute mark and the 40-minute mark with new constraints. If your initial design cannot absorb one new requirement without a major rewrite, your abstraction layer is too thin.

Round 5: Hiring Manager

The hiring manager round at Uber is split between technical and behavioral. The technical half is a thorough interrogation of the candidate's current or most recent project: architecture decisions, scaling challenges, what went wrong and how they fixed it, what they would redesign if starting over.

The behavioral half uses situational questions: "Tell me about a time you disagreed with a technical decision and what you did" or "Describe a production incident you owned and how you resolved it." The STAR format (Situation, Task, Action, Result) works here, but the interviewer probes for specifics. Vague answers ("I collaborated with the team to solve it") fail. Specific answers ("I identified the root cause was a missing index on the orders table, added it during a maintenance window, and p99 dropped from 4.2 seconds to 180 milliseconds") pass.

The candidate passed. The hiring manager round at Uber is not a gatekeeping round if you have cleared the technical rounds; it is a validation round. Strong technical performance plus a reasonable behavioral showing equals a pass.

Round 6: Bar Raiser

The bar raiser is the final round and the one with veto power. The interviewer is not from the hiring team. They have no stake in filling the position. Their job is to evaluate whether the candidate meets Uber's engineering bar independently of the hiring team's enthusiasm.

The format is a sustained cross-examination of the candidate's past project design decisions. Not "what did you build" but "why did you build it that way, what alternatives did you consider, and what would break if traffic doubled." The bar raiser is looking for engineering maturity: the ability to reason about trade-offs, acknowledge limitations in your own design, and articulate what you would change with more time or resources.

The candidate passed. The bar raiser round is where candidates with strong coding skills but shallow design thinking get caught. If you can solve a LeetCode Hard but cannot explain why you chose a hash map over a balanced BST in your last project, you will struggle in this round.

The candidate's preparation approach

Two named preparation areas from the source post:

1. LeetCode. Standard DSA preparation covering arrays, strings, trees, graphs, dynamic programming, and advanced data structures including Segment Trees.
2. Machine coding with concurrency and OO design. Practice problems involving thread safety, producer-consumer patterns, and incremental-requirement design.

The absence of system design course names is notable. At SDE-2 level, candidates who have worked on production systems for 2 to 5 years typically rely on their own experience for system design rather than courses. The candidate's system design performance (passing with schema modifications) suggests practical experience with distributed systems rather than course-level preparation.

How to prepare for Uber's 6-round gauntlet

Six takeaways from this specific 6-round loop.

1. Uber's SDE-2 loop is 6 rounds, not 4. Budget your preparation and your energy accordingly. The machine coding round and the PS round are independent evaluations. Strength in one does not compensate for weakness in the other.
2. Machine coding is the highest-signal round. The job scheduler with concurrency tests design under evolving requirements, correctness under concurrency, and coding speed simultaneously. Practice machine coding with a timer and a friend who adds requirements mid-session.
3. Segment Trees are fair game at SDE-2. If you have never solved a Segment Tree problem, add 10 to 15 problems to your preparation plan. Lazy propagation is the dividing line between "seen Segment Trees before" and "can use them under pressure."
4. System design rounds add requirements mid-round. Design your initial architecture with an extension layer. If adding "message search" or "encryption" requires a rewrite, your abstraction is too thin.
5. The bar raiser evaluates engineering maturity, not coding skill. Prepare to defend your past design decisions with trade-off reasoning. "We chose X because of Y constraint, and the alternative Z would have cost us W" is the shape of answer that passes.
6. Hints in the PS round are not automatic failures. Taking one or two hints and reaching a complete solution is a pass. The interviewer is testing whether you can incorporate new information, not whether you solved it cold.

What candidates ask about Uber SDE-2

What is Uber's L4 level equivalent to at other companies? Uber L4 is SDE-2, roughly equivalent to Amazon L5, Google L4, and Microsoft 62. It is the mid-career band for engineers with 2 to 5 years of experience.

How long did the full pipeline take from CodeSignal to offer? The source post does not specify the exact timeline. Based on Uber India's typical scheduling cadence, a 6-round loop usually takes 3 to 5 weeks from first screen to offer.

Was the machine coding round in Java, Python, or something else? The candidate's language choice is not disclosed. Uber allows any mainstream language for machine coding. Java and Python are the most common choices for concurrency problems due to their threading libraries.

Is the CodeSignal screening proctored? Yes. CodeSignal's screening environment monitors for tab switches and external tool usage. Treat it as a proctored exam.

How much does Uber SDE-2 (L4) pay in India? The candidate did not disclose compensation due to NDA. Uber India L4 compensation in 2025 typically ranges from ₹45 LPA to ₹70 LPA total, depending on team, location, and negotiation. This includes base salary, restricted stock units, and performance bonus.

Did the candidate have a referral? Not disclosed. The source post does not mention how the candidate entered the pipeline (referral, cold application, or recruiter outreach).

Where this comes from

This analysis is anchored on a post published on Roundz Substack with round-by-round detail for all 6 rounds. Original post: Interview Experience: Uber SDE-2 (L4), India, Offer

PapersAdda's verification standard requires a publicly identifiable source URL, per-round detail from the candidate, and an outcome statement. This post meets all three. The candidate described each round's format, topic, and difficulty, and confirmed the outcome as an offer. Compensation is NDA-restricted, which is noted transparently.

Related verified interviews

• The Amazon L5 comparison: Amazon SDE-2 (L5) 2026 Tier-3 engineer's real 5-round loop to ₹60L, 5 rounds at Amazon versus 6 at Uber for the same level band
• The SDE-1 version of this loop: Uber SDE-1 India 2025, all 3 OA problems solved, rejected at Google Sheets LLD, what happens when the machine coding round goes wrong at the entry level
• The Amazon SDE-1 onsite: Amazon SDE-1 onsite 2026 Tier-2 engineer's real 4-round loop, a shorter loop at the same tier of company
• The Microsoft new-grad loop: Microsoft new grad 2026 4-round interview real questions, a different company's approach to entry-level technical evaluation