Continued from Part 1. This part contains all 9 detailed candidate timelines and closes the September 2026 Amazon roundup.

Coverage: Candidate timelines 01–09, including five reports explicitly labeled SDE II, Software Engineer II, or Amazon II and four reports whose exact level was not disclosed, followed by preparation priorities and the final checklist.

9 detailed candidate timelines, loops, and screens

01: Offer — evolving BFS, return LLD, safest path, and config sync

Report date: September 13, 2026 · Level: Explicit SDE II · Result: Offer

OA

1. AI-assisted repository debugging.
2. One undisclosed DSA problem.

Round 1 — DSA + LP

1. Learn and Be Curious and Ownership; exact BQ wording was not published.
2. A custom BFS gained constraints until it became multi-source BFS.
3. The candidate completed the code but did not disclose a known missed edge case.

Round 2 — LLD + LP

1. Dive Deep and Invent and Simplify; exact BQ wording was not published.
2. Design and partly implement the return drop-off reservation system.
3. Follow-up: locate nearby eligible stores. The interviewer hinted at nearby-driver search.

Round 3 — Bar Raiser

1. Customer Obsession and Deliver Results; exact BQ wording was not published.
2. Solve a safest-path grid problem and optimize it repeatedly.

Round 4 — Hiring Manager + HLD

1. Earn Trust and Have Backbone; Disagree and Commit; exact BQ wording was not published.
2. Design multi-device configuration synchronization and handle added failure scenarios.

02: Downlevel — SDE II loop resulted in an SDE I offer

Report date: September 9, 2026 · Level: Explicit SDE II · Location: Bangalore · Experience: 3.3 years · Result: SDE I offer

OA

1. Binary-search-on-answer problem.
2. AI-assisted Python/Django debugging.

Phone screen

1. Count Stepping Numbers in Range, with an undisclosed follow-up.
2. Capacity to Ship Packages Within D Days.
3. Bias for Action; exact BQ wording was not published.

HLD/LLD scheduling mix-up

1. Design Amazon inventory management.
2. Follow-up on concurrent orders for the final unit.
3. Earn Trust; exact BQ wording was not published.
4. Feedback: Neutral.

DSA

1. Container With Most Water.
2. Follow-up: Trapping Rain Water.
3. Cheapest Flights Within K Stops.
4. Bias for Action and Dive Deep.
5. Feedback: Very Positive.

Hiring Manager + LLD

1. Deliver Results.
2. Design a distributed rate limiter.
3. Feedback: Positive.

Bar Raiser

1. Knapsack-style problem; greedy was challenged and the candidate pivoted to DP but did not finish.
2. Customer Obsession and Ownership.
3. Feedback: Negative.

Amazon did not clear the SDE II bar and instead offered SDE I.

03: Two openings, an unusual phone screen, and a withdrawn onsite

Report date: September 2026 · Level: Explicit SDE II · Location: Seattle team for the second opening · Result: Withdrew

First opening — OA

1. Drone-delivery loop problem using route travel times and prefix sums.
2. AI-assisted project debugging based on README function definitions.
3. The position was filled before the phone screen.

Second opening — phone screen

1. About 30 minutes of BQ and 30 minutes of coding.
2. BQ: missed deadline, self-initiated project, and high-impact project; all had undisclosed follow-ups.
3. Coding: schedule vehicle registrations as evenly as possible under recurring registration, holiday, and daily-hour constraints.
4. Follow-up: change the daily window from 10 a.m.–4 p.m. to 11 a.m.–3 p.m.
5. The candidate advanced despite believing the screen had gone poorly.

Onsite day one — Hiring Manager

1. Deep dive on an AI-agent project: implementation, users, and departments.
2. BQ: biggest mistake, process improvement, conflict, and criticism.
3. Coding: Meeting Rooms. The candidate initially forgot to retain the maximum active overlap during the sweep, then corrected it.

Onsite day one — second interviewer

1. BQ: new learning applied at work, greatest company benefit, and customer satisfaction. The candidate reported no follow-ups.
2. Design a low-latency notification-ingestion microservice for burst traffic.

The candidate canceled the remaining next-day interviews and withdrew.

04: Palindrome phone screen

Report date: Approximately September 6, 2026 · Interview date: August 29, 2026 · Level: Explicit Software Engineer II / SDE II · Experience: Four years · Result: Passed phone screen

1. Three behavioral questions focused mainly on Ownership and Dive Deep. The exact questions and follow-ups were not preserved.
2. Solve a relatively straightforward palindrome-based coding problem.
3. Explain the approach, edge cases, and time/space complexity.
4. The recruiter confirmed the pass the following day.

05: Three technical prompts before LPs

Report date: Approximately September 18, 2026 · Level: Explicit “Amazon II” in the source title · Experience: Three years · Status: Not disclosed

Technical portion first

1. LeetCode 146, LRU Cache. The candidate solved it quickly and received two undisclosed follow-ups.
2. Design a robot restaurant with many robots and explain how they should coordinate. The broad prompt led to many follow-ups.
3. Design a Kafka-based file-watcher system that can handle millions of files.

LP portion

1. Tight deadline: was there a project that could not be completed because of time?
2. Dive Deep: describe finding the root cause of a difficult problem; the candidate framed it as a debugging story.

06: Resilient Hub screen — ownership, AI impact, and timestamped values

Report date: September 2026 · Level: Not disclosed · Result: Rejected

1. Two interviewers participated: a product manager and an SDE.
2. BQ first: a time the candidate took responsibility/ownership, and a time the candidate faced a challenge.
3. Follow-ups focused on AI usage, impact, number of users, and the manager’s reaction.
4. Coding: implement a latest-timestamp key-value object supporting timestamped set operations and get(key) returning the value at the maximum timestamp for that key.
5. The candidate spent too long understanding the prompt and did not complete the implementation.

The source describes the object as “LRU-like,” but no capacity or eviction behavior was reported. It is better understood as a latest-timestamp key-value store.

07: Three LPs left 15 minutes for Maximum Frequency Stack

Report date: Approximately September 1, 2026 · Level: Not disclosed · Experience: Ten years · Result: Rejected

1. The screen began with three standard Amazon LP questions; their wording was not preserved.
2. With roughly 15 minutes left, the interviewer asked LeetCode 895, Maximum Frequency Stack.
3. The candidate recognized a stack-of-stacks direction but provided an O(log n) approach and did not finish coding.
4. The interviewer later pointed out a correctness issue, and the rejection arrived the next day.

08: Four-round onsite with follow-up-heavy coding and design

Report date: Approximately September 20, 2026 · Level: Not disclosed · Experience: Four years · Application: Referral · Status: Source metadata says Pass; narrative said awaiting the result

1. Four onsite rounds with four interviewers.
2. Coding questions were described as ordinary LeetCode medium problems, but interviewers kept adding follow-ups.
3. System design repeatedly drilled into why each choice was made, then scale, consistency, and failure handling.
4. LPs were described as standard Amazon questions tied to prepared project stories.

The source did not preserve the individual coding or design prompts. The result field and narrative status do not fully agree, so the outcome remains qualified rather than treated as an offer.

09: First-round screen with conflict and tight-timeline BQ

Report date: Approximately September 3, 2026 · Level: Not disclosed · Experience: Four years · Application: Referral · Status: In process when reported

1. The candidate had completed only the first round when posting.
2. Preserved BQ topics were conflict and a tight timeline.
3. The screenshot does not preserve the technical prompt, LP follow-up wording, or outcome.

What repeated often enough to prioritize

Tier 1: Practice until automatic

1. Follow-up-driven coding. Medium problems were repeatedly extended through new constraints, required optimizations, or implementation details.
2. Failure-aware design. Scale, consistency, retries, concurrency, and failure handling appeared across several independent reports.
3. Detailed LP evidence. Ownership, Dive Deep, Customer Obsession, deadlines, conflict, mistakes, and personal impact repeatedly appeared.

Tier 2: Be ready to explain deeply

1. Cache and timestamp semantics. Know LRU internals, frequency stacks, and latest-timestamp lookup.
2. Concurrency. Prepare overselling, locking, transactions, multi-instance rate limiting, and safe updates.
3. Correctness after apparent acceptance. One candidate corrected a sweep-line answer only after interviewer pressure.

Tier 3: Current-cycle differentiators

1. AI-assisted engineering tasks. Practice reading an unfamiliar repository, verifying AI suggestions, and debugging from documentation rather than expecting the assistant to reveal the fix.
2. AI impact in BQ. One report’s follow-ups asked how AI was used, how many people adopted it, what impact it produced, and how management reacted.