All experiences
Offered2026-08-09·4 min read

Amazon SDE-2 — 4 Rounds, Offer Accepted

3 YOE from quick commerce (Swiggy, Zepto). DM'ed a recruiter, got HackerRank link in 3 days. System design, HM/LLD discussion, DSA, and bar raiser. Offer: 46L base + 29L joining + 38L ESOPs.

Company

Amazon

Role

SDE-2

Rounds

4

YOE

3

Timeline

  • DM'ed recruiter: looking for openings
  • HackerRank link: 3 days after DM
  • Recruiter follow-up: ~2 weeks after OA
  • Interview scheduling: 10 more days
  • All 4 rounds: within a week
  • Offer: 2 days after bar raiser

Background

  • 3 YOE at quick commerce startups (Swiggy and Zepto)
  • Used Golang primarily
  • No prior FAANG experience

Round 1 — Online Assessment (HackerRank)

2 DSA questions (both Medium):

  1. Binary search based — standard medium
  2. Trees — traversal/manipulation problem

Followed by LP-based multiple choice questions (behavioral scenarios with choice-based answers).

Nothing tricky here. Standard OA.

Round 2 — System Design (60 min)

Problem: Design an auction platform.

Interviewer focused on:

  • Database schemas (how to model bids, auctions, users)
  • Scaling for high concurrency (many bids at once)
  • Low latency API design (caching layer)

No LPs asked. Instead, asked me to explain an interesting project from past work.

Follow-ups asked:

  1. How do you handle edits to bids? (I suggested allowing multiple bids rather than edit — cleaner audit trail)
  2. How would you support a near real-time leaderboard? (Redis sorted sets)
  3. How would you identify/track the winning bid? (Can't rely on Redis as source of truth — need DB confirmation)

Prep that helped: Hello Interview. Hadn't solved this exact question before but the framework was familiar enough.

Verdict: Passed with good feedback.

Round 3 — HM / LLD Discussion (60 min)

This was supposed to be an LLD round but ended up being mostly project discussion:

  • Deep dive into my work at Zepto — architecture, design choices
  • "Could you build the inventory service using NoSQL instead of SQL?" — trade-off discussion
  • Last 10 minutes: what design patterns I used in past projects (explained Strategy and Observer patterns from real code)

No formal LLD problem. No LPs asked.

Verdict: Positive feedback. I was worried because no actual LLD question was asked — but turns out the project discussion WAS the evaluation.

Round 4 — DSA (60 min)

Interviewer was a younger SDE-2.

Questions:

  1. Implement browser history navigation — forward, back, visit new URL. Used a doubly linked list / two-stack approach.
  2. *Implement basic arithmetic (+, -, , /) without inbuilt functions — bit manipulation for addition, repeated addition for multiplication, etc.

I coded in Golang. The interviewer wasn't familiar with Go and I had to explain syntax basics. Felt like the interviewer wasn't able to ask proper follow-ups because of the language gap.

Verdict: Positive but slightly hesitant.

Bar Raiser (60 min)

LPs: Focused heavily on Ownership. Needed strong STAR stories demonstrating end-to-end ownership of features/systems.

DSA Questions:

  1. Rotten Oranges — multi-source BFS on grid. Classic.
  2. Asteroid Collisions — stack-based simulation.

Both are well-known problems. Speed and clean implementation mattered.

Verdict: Passed.

The Offer

Heard back in 2 days:

  • Base: ₹46L
  • Joining bonus: ₹29L (split over 2 years)
  • ESOPs: ₹38L (4-year vest)

Tips for Amazon SDE-2

  1. DM recruiters directly — it works. Be specific about the role you want.
  2. System design prep: Hello Interview + watching mock interviews. They focus on DB schema, caching, and scaling — not just drawing boxes.
  3. LPs are unpredictable — some rounds ask them heavily, some don't ask at all. Prepare 8+ stories anyway.
  4. HM round might not have a formal LLD problem — your project deep dive IS the evaluation. Know your past work intimately.
  5. Language choice matters slightly — if you use a less common language (Go, Rust), the interviewer might not be able to probe as deeply. Double-edged sword.
  6. Bar raiser DSA is medium difficulty — but you need to solve FAST with clean code. They're calibrating speed + quality, not just correctness.

Join 500+ engineers

Weekly interview questions + comp data insights. No spam. Unsubscribe anytime.

Powered by Substack. Your email is safe.