🧪 My Amazon L5 (SDE II) Interview Experience – What Went Well and What to Avoid
- Location
- India
- Experience
- 3.5 years
- Posted
- 22 May 2025
Summary
Recently interviewed for Amazon's SDE II (L5) role and made it through 4 rounds. I did well in DSA and Leadership Principles, but got a mixed review on System Design, which ultimately led to a rejection. Sharing this so others can learn from it.
How it went
1️⃣ Round 1 – HLD + LP (23rdApril)
- Type: High-Level System Design
- Question: "Design your current AI chat system"
- I explained the architecture I had built in my current role – real-world experience helped.
- Walked through:
- Client → Gateway → Services → Database
- Real-time messaging via WebSocket
- Stateless backend, rate-limiting, vector DB usage, etc.
- LPs: Went well — interviewer mentioned "structure can improve" in LP answers.
- Feedback: Interviewers were impressed technically. Small feedback on improving LP structure.
2️⃣ Round 2 – DSA + LP (24thApril)
- Type: Graph + Heap
- Question 1: Find a connecting path between flights (flight booking system scenario).
- Follow-up 1: Add weights (costs) → used Dijkstra’s Algorithm.
- Follow-up 2: Handle
to → fromdirection → made bidirectional graph changes. - Used heap-based implementation, handled edge cases.
- LPs: Focused on handling ambiguity, delivering results, taking ownership.
- Feedback: Positive. One of my strongest rounds.
3️⃣ Round 3 – LLD + LP (2nd May)
- Type: Low-Level Design + Deep Dive
- Question: Design Splitwise:
- Users create groups
- One user pays, others split bill
- View balances and settle-up
- I built the class diagram, schema design, and balance tracking logic.
- I missed stating functional and non-functional requirements explicitly up front.
- I assumed the requirements were already clear, so I jumped into design quickly.
- LPs: Went fine — no negative signals.
- Feedback (post-interview): Mixed. Interviewer expected more upfront structure and deeper reasoning on how to handle things like balance tracking, consistency, and settlements.
4️⃣ Round 4 – Bar Raiser (LP + DSA) (19th May)
- LPs: 40 minutes — covered Invent & Simplify, Ownership, Bias for Action.
- Focused on scaling an app I had worked on — taking DAUs from 1k → 25k.
- DSA: Stack-based problem – Asteroid Collision (LeetCode Medium).
- Solved in under 10 mins with optimal solution.
- Follow-up: “How would you optimize memory usage?” → discussed.
- Feedback: Strong positive.
❌ Final Verdict
- Coding: Strong hire
- LPs: Acceptable/good
- System Design: Mixed feedback → led to rejection
Questions asked (4)
- 01
Design AI Chat System
System Design
Design your current AI chat system.
- 02
Flight Path with Weights and Bidirectional Edges
Data Structures & Algorithms
Find a connecting path between flights (flight booking system scenario). Follow-up 1: Add weights (costs). Follow-up 2: Handle
to → fromdirection. - 03
Design Splitwise
System Design
Design Splitwise with the following features: Users can create groups, one user pays and others split the bill, users can view balances and settle up.
- 04
Asteroid Collision
Data Structures & Algorithms · Medium
A stack-based problem, LeetCode Medium. Follow-up: How would you optimize memory usage?
How they prepared
🧠 Learnings (TL;DR)
-
System Design: Always Start With Checklist
Even if the problem looks simple or familiar:- Functional + Non-Functional Requirements
- Constraints, Assumptions, User flows
- Only then move to class diagrams/schema
-
Depth Is Key
Amazon values clarity in thought. Instead of racing into the solution, pause to:- Talk through trade-offs
- Consider how the system handles scale, consistency, and failure
-
LPs Need Structure
- Use STAR: Situation, Task, Action, Result
- Expect follow-up — they will test consistency
-
Coding Interviews Are Only One Part
Even if you solve all DSA problems optimally, design rounds carry equal or more weight at L5. -
Bar Raiser Will Probe Hard on LPs
Be authentic, but also tight. Keep your story focused and measurable.