What topics are tested in Amazon SDE-2 interviews?

Amazon SDE-2 interviews test: Heaps, Graphs, Binary Search, MST, LP: Judgment, Arrays, HashSets, Greedy.

Does Amazon SDE-2 have a system design round?

Yes, Amazon SDE-2 includes a system design round.

Amazon

Q: How many rounds are there in Amazon SDE-2 interviews?

Amazon SDE-2 typically has 5 rounds. Amazon SDE-2 (L5) interviews run a 4–5 round virtual loop: two coding rounds, a system design round, and a Bar Raiser round. LP stories at this level must reflect ownership of a full feature or service — not individual tasks. The system design round is graded partly through an LP lens: interviewers expect you to mention Frugality and Customer Obsession unprompted when discussing trade-offs. The Bar Raiser is from a different org and has veto power regardless of how the other rounds go.

E-commerce / Cloud · Seattle, WA

Amazon operates the world's largest marketplace and AWS, the dominant cloud platform powering 33% of the internet.

Related Companies

Rounds: 5 rounds
Focus: DSA + System Design + LP
System Design: Yes
Timeline: 3–4 weeks to offer

Role:

Rounds 1–2 are fully visible · Round 3+ unlocked with a free account

Online Assessment

Round 1 of 5OA

90 min

HeapsGraphsBinary SearchMSTLP: Judgment

Two harder coding problems than SDE-1 OA, plus the Work Simulation and Work Style Survey. For SDE-2 candidates, the OA coding problems lean Medium-Hard and the Work Simulation scenarios involve system-level judgment — prioritizing a P1 incident while managing a sprint commitment, or deciding how to respond when a teammate's data contradicts yours. Amazon's OA environment disables the debugger — your code needs to run clean on first submission.

Medium

What to expect

Coding Problem 1 (Medium-Hard): e.g., 'Given a stream of package weight readings, return the median after each reading' — two-heap approach expected
Coding Problem 2 (Hard): e.g., 'Minimum cost to connect all fulfillment centers' (Prim's or Kruskal's MST) — graph problems at this level require knowing which algorithm to reach for and why
Work Simulation: inbox scenarios include a P1 production alert landing during a sprint planning meeting — they want to see incident response instinct, not panic
No partial credit — a solution that passes 7/10 test cases scores the same as 0/10; budget time to verify edge cases before submitting

Practice on KnowEase.AI

DSA Practice — Heaps →DSA Practice — Graphs →

Coding Round 1

Round 2 of 5DSA / Coding

60 min

ArraysHashSetsGreedyLP: OwnershipLP: Highest Standards

LP questions followed by 1–2 coding problems. At SDE-2, LP stories must show ownership of a complete feature or service, not a single bug fix or task. Sliding window, hash map, and union-find problems are consistently reported in this slot. The interviewer probes whether you can generalize your solution — expect the problem statement to change mid-round.

Medium

What to expect

LP: 'Describe the largest end-to-end technical project you personally owned — from design through production (Ownership)' — scope matters; owning a ticket queue doesn't cut it here
LP: 'Tell me about a time you raised the engineering quality bar for your team (Insist on the Highest Standards)' — code review culture, test coverage initiatives, and on-call process improvements are all fair game
Coding: e.g., 'Longest consecutive sequence in an unsorted array' (O(n) with HashSet, no sorting) or 'Task Scheduler — minimum CPU intervals to finish all tasks' (greedy + frequency counting)
Follow-up: 'What if the array has duplicates?' or 'What if task priorities change dynamically?' — missing the follow-up costs more than missing the initial solution

Practice on KnowEase.AI

DSA Practice — Arrays →

Coding Round 2

Round 3 of 5DSA / Coding

60 min

HeapsTwo HeapsDP State MachineLP: Learn and Be Curious

Harder coding problem with LP focus on Learn and Be Curious. At SDE-2, interviewers expect you to proactively discuss the trade-offs between your solution and alternatives you considered — not just give the optimal answer. Two-heap, DP state machines, and graph problems appear here. Passing the coding problem but failing to articulate trade-offs is a common rejection path at this level.

What to expect

LP: 'Tell me about a technology or system concept you taught yourself in the past year to solve a real problem (Learn and Be Curious)' — citing a paper, a new AWS service, or a pattern from a system you read about all work; 'I watched a YouTube tutorial' does not
Coding: e.g., 'Find the median of a data stream' (two heaps: max-heap for lower half, min-heap for upper) or 'Maximum profit from stock with cooldown' (DP state machine: hold / sold / rest states)
Follow-up: 'What if 5% of your data points are corrupt negative values — how does that affect your median approach?'
Trap: failing to mention that O(log n) per insert with two heaps is preferred over O(n) re-sort — SDE-2 candidates are expected to lead the complexity conversation, not wait to be asked

Full details unlocked with a free account

System Design

Round 4 of 5System Design

60 min

Pub/SubCDNMessage QueuesDatabasesSLAsLP: Frugality

Design a large-scale Amazon-domain system. The most frequently reported prompts in 2024–2025 involve Amazon's own infrastructure: order notification pipelines, Prime Video content delivery, and product search indexing. What makes Amazon's design rounds distinctive is that the interviewer evaluates LP alignment alongside technical choices — you are expected to invoke Frugality and Customer Obsession when discussing trade-offs, unprompted. 'I'd use SQS over Kafka here for Frugality — the cost-to-throughput ratio at this scale favors managed queues' lands better than a technically correct answer with no LP framing.

What to expect

e.g., 'Design Amazon's order notification system' — SNS fan-out to SQS, email/SMS channels, at-least-once delivery guarantee, DLQ for failed notifications, idempotency on retries
e.g., 'Design Amazon Prime Video's content delivery pipeline' — CDN tiering, adaptive bitrate streaming, regional cache invalidation on new uploads, failure handling when an edge node goes dark
Frugality angle: 'How do you reduce delivery cost at 1B notifications per day?' — spot instances for batch, reserved capacity for real-time, tiered storage for cold logs
Customer Obsession angle: 'What SLA do you commit to for order confirmation delivery and how do you detect SLA breach before the customer notices?'
Trap: designing a technically sound system with no discussion of failure modes — Amazon interviewers explicitly probe 'what happens when SNS drops 0.1% of messages?' and expect a concrete answer

Full details unlocked with a free account

Bar Raiser

Round 5 of 5Bar Raiser

60 min

LP: All 16 PrinciplesSelf-AwarenessJudgmentTechnical Depth

An independent interviewer from a completely different Amazon org. Their job is to evaluate whether you raise the bar above the current SDE-2 median — not whether you're good enough for this specific team. Amazon revived Bar Raiser involvement at lower levels in 2024; at SDE-2 it's standard. The Bar Raiser can veto a hire even if every other interviewer voted yes. They probe LP stories until the underlying thinking is exposed — polished STAR narratives tend to collapse under their 5-why follow-ups.

What to expect

5-why follow-ups on LP stories: 'You chose DynamoDB — walk me through the alternatives you considered and why you rejected them', then 'What would have to be true about your access patterns for that to be the wrong call?'
Self-awareness probe: 'Tell me about a technical decision you made that you'd make differently today — and specifically what changed in your thinking'
May include a fresh coding or design problem not seen in any prior round — they're evaluating structured thinking under pressure, not whether you recognize the problem
Trap: over-polished STAR stories with no rough edges — Bar Raisers are trained to detect rehearsed answers and will probe until the story either holds up or falls apart

Full details unlocked with a free account

Questions — coming soon.

Company Guide — coming soon.

About Amazon SDE-2 Interviews

Amazon SDE-2 (L5) interviews run a 4–5 round virtual loop: two coding rounds, a system design round, and a Bar Raiser round. LP stories at this level must reflect ownership of a full feature or service — not individual tasks. The system design round is graded partly through an LP lens: interviewers expect you to mention Frugality and Customer Obsession unprompted when discussing trade-offs. The Bar Raiser is from a different org and has veto power regardless of how the other rounds go.

See interview loops for other companies:

Microsoft Uber Flipkart Razorpay Walmart Coupang Rippling Rubrik Target

Start your Amazon SDE-2 prep plan

Practice the exact topics tested in each round — DSA, system design, SQL, and LLD.

Start free →