Amazon · Phone screen → Onsite
In processAmazon New Grad SDE Onsite Interview Questions: 17 Experiences
Amazon new grad SDE onsite loops, Apr–Sep 2026: 17 reports where GenAI questions joined LP, LLD and medium coding in almost every loop.
At a glance
- Process: a virtual or in-person final of three or four back-to-back rounds, each pairing a story with coding or LLD, often plus a behavioral-only Bar Raiser (04, 06, 07, 09, 12, 15).
- Asked recently: GenAI questions in 13 of 17 SDE reports; an LLD or OOD round in 13; LRU caches and Meeting Rooms II twice each (01, 12; 05, 18).
- Difficulty: the hard parts were spotting OOD requirements early (02) and deep story follow-ups (05, 06, 07).
- Practise first: LRU Cache for Query Results, Meeting Rooms II, Movie Theater Ticket Pricing and Package Dependency Order.
Coverage: 17 Amazon SDE I new grad candidates, post dates April 25 to September 4, 2026 (United States, Seattle, Vancouver, unstated). Offers in 07 and 12; five rejections. A Front End Engineer new grad loop is covered in the frontend roundup instead.
What came up
Counts are out of the 17 SDE reports; one candidate can sit in several rows.
| Topic | Reports | Last seen |
|---|---|---|
| Behavioral and Leadership Principles | 16 of 17 | Sep 2026 |
| GenAI and AI-use questions | 13 of 17 | Sep 2026 |
| LLD and OOD | 13 of 17 | Sep 2026 |
| Heaps, Top K, intervals and streams | 9 of 17 | Sep 2026 |
| Trees, graphs and linked lists | 9 of 17 | Aug 2026 |
| Hash maps and strings | 8 of 17 | Sep 2026 |
| Caches and rate limiters | 4 of 17 | Sep 2026 |
| Greedy and binary search | 3 of 17 | Aug 2026 |
| Dynamic programming | 2 of 17 | Sep 2026 |
Questions we found
Practice links are in the entries.
Coding
| Question | Reports | Last seen |
|---|---|---|
| LRU Cache (query-result cache) | 2 | Sep 2026 |
| Meeting Rooms II | 2 | Aug 2026 |
Each of these was named once (entry in brackets):
- Heaps, intervals and streams: Sliding Window Maximum (02), Top K Frequent Elements (03), Find Median from Data Stream (04), catalog Top K (06), rate limiter (06), Merge Intervals variant (07), Top K slowest logs (14), least-loaded server (15), Meeting Rooms (18).
- Trees, graphs and linked lists: relationship between two people (01), linked-list queue (04), tree with height, DFS and LCA (05), Path Sum II (07), Binary Tree from Descriptions, LC 2196 (12), Package Dependency Order (14), break a linked-list cycle (16), Circular Student Arrangements and siblings or cousins (17), multi-source BFS (18).
- Hash maps, strings, DP and greedy: Buy/Sell Stock with Cooldown (01), merge
kdictionaries (04), spreadsheet cell dependencies (06), First Non-Repeating Character (07), first user seen once (08), fewest reward points and Reorganize String (09), iterator (03), Valid Parentheses and Two Sum (11), Group Anagrams, LC 49 (14), two-day loyal customers (15), minimum cash pieces and Caesar Cipher (16), Longest Consecutive variant (17).
Low-Level Design (LLD / OOD)
Each named once: Train Route Fare Calculator (02), shopping cart (03), Pizza (05), AWS billing service (07), playlist generator (08), movie ticket pricing (09), Unix file search API (12), GenAI chatbot architecture (11), parking lot (13), log rate limiter (15), publish/subscribe (17), deck of cards (18).
Behavioral (BQ)
| Question | Reports | Last seen |
|---|---|---|
| GenAI use, wrong output, validation | 13 | Sep 2026 |
| Conflict or a difficult teammate | 4 | Sep 2026 |
| Work outside your role | 4 | Aug 2026 |
| A difficult decision or trade-off | 3 | Sep 2026 |
| Finding a root cause | 3 | Aug 2026 |
| Negative feedback or criticism | 3 | Aug 2026 |
| A tight deadline | 3 | Aug 2026 |
| An obstacle or blocker | 3 | Aug 2026 |
| Leaving your comfort zone | 2 | Aug 2026 |
| Ambiguity or missing data | 2 | Aug 2026 |
| What you would do differently | 2 | Aug 2026 |
| Short-term sacrifice | 1 | Aug 2026 |
Also asked: a team heading the wrong way, disagreeing with the team's direction, collaborating outside your area, an improvement nobody asked for, learning a technology for a real project, the hardest part of a recent project, which project best shows how you work, an achievement not yet covered, a mistake with real consequences, Why Amazon, when GenAI is and is not a good fit, and making a GenAI feature more reliable.
17 SDE new grad loops
01: LRU Cache and a stock DP
Post date: September 4, 2026 · Role: SDE I New Grad · Location: United States
Four rounds were planned; one was canceled.
- LP, then LRU Cache with extra inputs and extensions.
- LP, then a data structure for the relationship between two people (model not given).
- GenAI, then Best Time to Buy and Sell Stock with Cooldown, in a round expected to be behavioral only.
02: Train fares, strings and a sliding window
Post date: September 3, 2026 · Result: Rejected
- Train Route Fare Calculator: on
G → U → H → K → Ieach segment costs 1, viacalculateFare(start, stop). Follow-ups: stations and routes as objects, capped weekend fares. Spotting the OOD needs early was the hard part. - A string or substring problem with tests, including the empty string, plus GenAI.
- Behavioral: difficult teammates and decisions, with detailed follow-ups.
SlidingWindow(k)withput(number)andgetMax()over the lastkvalues, solved in O(k); follow-ups compared arrays, hash maps, linked lists, heaps and a monotonic queue. Practice: Sliding Window Maximum and the Sliding Window Maximum API.
03: Top K, a shopping cart and an iterator
Post date: August 31, 2026
Three video interviews of about an hour each:
- A Top K frequency problem.
- OOD for a shopping cart (requirements not shared).
- Implement an iterator.
04: An in-person loop in Vancouver
Post date: August 31, 2026 · Location: Vancouver · Cohort: NG26
Amazon paid flights, a hotel for two days, 70 a day for meals and 100 for transport (currency not given).
- GenAI (past projects, fit, uses, reliability), then LLD: a linked-list queue with O(1) enqueue and dequeue without built-in structures, plus
delete(x)andremoveAllDuplicates(). - Two LP questions, then merge two dictionaries with loops and extend to
k(merge rules not shared). - A behavioral-only Bar Raiser, answered with internship stories.
- One BQ, then Median of a Data Stream framed as an API (LLD version).
05: Trees, a stack, pizza and meeting rooms
Post date: August 28, 2026
- A tree problem (height, DFS, LCA), a monotonic stack, Pizza OOD, and meeting rooms with a priority queue; exact prompts not shared. Related: Lowest Common Ancestor Implemented with a Stack and Meeting Rooms II.
- Behavioral: common LP topics and AI output; follow-ups checked that each story showed the principle.
06: Rate limiter, Top K and a spreadsheet
Post date: August 27, 2026 · Result: Rejected
- A rate limiter, similar to Logger Rate Limiter (LC 359) and Design Hit Counter (LC 362).
- AI usage, then a catalog Top K with memory optimization (close to LC 215 and Kth Largest Element in a Stream, LC 703).
- A behavioral-only Bar Raiser on missing data, conflict and root cause, probing decisions, alternatives, evidence and results.
- Spreadsheet cell dependencies; LC 631 and LC 3484 are similar but not exact.
07: An AWS offer in Seattle
Interviews: August 20 and 24, 2026 · Post date: August 29, 2026 · Location: Seattle (AWS) · Result: Offer on August 27
- A blocker and what they would do differently, then Path Sum II and a harder variation; hints helped on the follow-up.
- Work outside normal responsibilities, then an AWS billing service for one user (services, usage, rates), on classes, trade-offs and changing needs.
- About 30 minutes of GenAI, then First Non-Repeating Character and a Merge Intervals variation.
- A behavioral-only Bar Raiser on decisions and rejected alternatives.
08: Logs, DP and a playlist in Vancouver
Interview: August 24, 2026 · Post date: August 29, 2026 · Location: Vancouver · Result: Rejected three days later
Three onsite rounds and one virtual, same day.
- From time-ordered access logs, find the earliest user who appears exactly once, then complexity and LPs.
- A DP problem, then everyday AI use and a real AI hallucination.
- LPs, projects and learning, then a playlist-generator LLD with many follow-ups (related practice).
- BQs on conflict, leaving a comfort zone, recent projects and their difficulties.
09: Reward points, a string and ticket pricing
Post date: August 21, 2026 · Result: Passed; no offer update yet
- A behavioral-only Bar Raiser.
- Pay with the fewest reward points: each program has a balance and a changing cash value. Example: cover $168 with 36,000 airline points at $0.012, 12,000 store points at $0.009 and 25,000 card points at $0.0085; fail if impossible.
- Rearrange a string so no two neighbors match; return any valid answer or failure.
- Extensible movie ticket pricing:
Price = (base + extras) × showtime × seat tier, where format sets the base; new options must fit without big changes.
11: A phone interview with GenAI and Two Sum
Post date: August 18, 2026
- Behavioral: a tight deadline, criticism and ambiguous requirements.
- GenAI: AI experience and chatbot architecture (Stateless Generative AI Chat Service is related practice).
- Coding: Valid Parentheses, Two Sum and a Two Sum variation.
12: File search, a cache and a tree, then an offer
Interviews: July 20 and August 12, 2026 · Post date: August 17, 2026 · Result: Offer on August 17
VO survey July 8, three rounds July 20, Bar Raiser August 12.
- Behavioral only: Dive Deep, negative feedback and GenAI.
- Sacrificing short-term gain, then a Unix file search API for files under a directory, adding minimum size, AND filters and symlink safety (LLD version).
- An obstacle, then an LRU query-result cache for a multi-time-zone database with traffic peaks; follow-ups on
get()cost, concurrency, sequential IDs and rising traffic. - Leaving a comfort zone, then Create Binary Tree from Descriptions (LC 2196); follow-ups on the hash map, complexity, and out-of-order input that first forms a forest.
13: Parking lot, graphs and binary search
Post date: July 27, 2026 · Result: Pending
- LP, resume and a project deep dive.
- Two medium-to-hard questions on linked lists and graphs, both finished.
- Parking lot LLD (Multi-Floor Parking Lot is related practice).
- Hiring manager and GenAI questions on AI use, then a binary-search-on-the-answer problem.
14: A four-round VO
Interview: July 23, 2026 · Post date: August 14, 2026 · Result: Rejected
Applied in January; the VO invite came July 7.
- A GenAI story about unexpected output, then the Top K slowest log entries with a priority queue (similar to LC 636).
- Work outside their area, then a package dependency manager that must return a valid install order using topological sort.
- Trade-offs and a challenge, then Group Anagrams (LC 49) with a hash map and a bucket-based optimization.
- Behavioral: a tight deadline and a teammate conflict.
15: An in-person final: servers, customers and rate limiting
Interview: first week of July 2026 · Post date: July 8, 2026 · Role: SDE I New Grad · Location: United States · Result: Positive, but no offer because of business needs
- LPs, then assign partitions to the least-loaded server with a min heap, compared with a greedy option.
- LPs, then from two days of logs find customers present both days with at least two unique products, using nested dictionaries and sets.
- A behavioral-only Bar Raiser.
- An extensible log rate limiter with per-message and global policies; the candidate fixed a design where a policy changed state even when another rejected the event.
16: Cash change, a cycle and a cipher
Interview: June 18, 2026
Four back-to-back rounds:
- Extra responsibility and a commitment at risk, then the fewest cash pieces from $20, $10, $5, $1, 25¢, 10¢, 5¢ and 1¢ (6.35 = $5 + $1 + 25¢ + 10¢); follow-ups added early exit and limited inventory (Minimum Cash Pieces with Limited Inventory).
- Deep analysis, influencing a disagreement, improving after feedback and applying it later.
- Detect and break a linked-list cycle by setting the bad
nexttoNULL; follow-ups on Floyd's pointers, correctness, the entry node and a hash table. GenAI: an efficiency use and validating output; the resume deep dive covered memory ordering for IPC. - The hardest work, their own contribution and debugging, then a Caesar Cipher with a precomputed alphabet: keep case, wrap letters, skip other characters.
17: Permutations, tree relatives and pub/sub
Interview: May 13, 2026 · Post date: May 14, 2026
- Behavioral questions and follow-ups only.
- Print every ordering of
Nstudents, then in a circle (Circular Student Arrangements); then whether two tree nodes are siblings, cousins or neither. - Resume, then a Longest Consecutive Sequence variant where neighbors may differ by less than
k. - Behavioral, then OOD for an in-memory publish/subscribe system.
18: BFS, meeting rooms and a deck of cards
Post date: April 25, 2026 · Result: Rejected
- GenAI experience, then a BFS from several starting points.
- Obstacles, juggling tasks on a tight deadline, and revisiting a solution's trade-offs; then Meeting Rooms and Meeting Rooms II.
- Proactive work outside their role, then shuffle and sort a deck of cards in place: Clubs, Diamonds, Hearts, Spades, Ace through King.
Front End Engineer new grad
Entry 10, a Front End Engineer new grad loop for AWS EC2 Pricing (July–August 2026), is also covered in Amazon frontend engineer OA and interviews, entry 01.
Patterns across the reports
- GenAI is standard. 13 of 17 SDE loops asked about it (01, 02, 04–08, 11–14, 16, 18), up to about 30 minutes (07).
- A behavioral-only Bar Raiser. Entries 04, 06, 07, 09, 12 and 15 had one round with no coding.
Common questions
How many rounds is the Amazon new grad final loop?
Usually three or four back-to-back rounds (01, 03, 04, 08, 12, 14, 16), sometimes split across days (07, 12).
Does Amazon ask GenAI questions to new grads?
Yes, in 13 of 17 SDE reports: how they use AI, wrong output, and checking it (05, 08, 14, 16).
Is there system design in the Amazon new grad interview?
Mostly LLD or OOD instead: 13 of 17 loops had one, such as fares, billing or ticket pricing (02, 07, 09).
Keep practicing
- LRU Cache for Query Results (01, 12), Meeting Rooms II (05, 18) and Package Dependency Order (14).
- Amazon question banks (sign-in needed): coding, low-level design and behavioral.
- Amazon interview guide and Amazon hiring insights.
- Experience feed: more interview reports.
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.