AMD · Phone screen → Onsite
In processAMD Experienced Full-Time Interviews by Role through Oct 2026: 16 Experiences (Part 2 :)
Note: This is the most complete collection of interview information I could find in this round of searching. If friends need more information in the future, I may start another round of research. Good luck with your interviews! — 11-02-2026 (jane)
In this series
- Part 1: Senior Software System Designer; GPU Software Engineer; Senior Software Engineer - uProf; Compiler / Performance; Mid-level Software Engineer.
- Part 2 (current): SDE2; Senior Software Engineer - AI; MTS Software Development Engineer - GPU; Role not reported — post-offer start delay; Software System Designer 2 — Test Automation; Verification — PCIe; Role not reported — C and memory panel.
- Part 3: GPU team — exact role not reported; MTS Verification; Member of Technical Staff - Performance.
Coverage: 7 of 16 AMD Experienced Full-Time accounts in 7 separate role sections. One entry is a post-offer account with no disclosed technical questions.
Other AMD hiring levels
- Hiring Level Not Reported: 9 accounts.
- Intern / Co-op: 18 accounts in 2 parts.
- New Grad: 11 accounts in 2 parts.
Roles in this article
- SDE2 — 1 account.
- Senior Software Engineer - AI — 1 account.
- MTS Software Development Engineer - GPU — 1 account.
- Role not reported — post-offer start delay — 1 account.
- Software System Designer 2 — Test Automation — 1 account.
- Verification — PCIe — 1 account.
- Role not reported — C and memory panel — 1 account.
SDE2
1 account in this role.
What came up
Reported topics: coding tasks.
Questions we found
Algorithms / Data Structures
| Question | Frequency | All last seen dates |
|---|---|---|
| Count stair-climbing paths using one- or two-step moves | 1 | Reports: 2024-03-31 (report); Additional same-company practice history: 2026-09-06 |
| Implement a queue using stack(s) | 1 | Reports: 2024-03-31 (report) |
| Trapping Rain Water | 1 | Reports: 2024-03-31 (report) |
| Check whether a binary tree is height-balanced | 1 | Reports: 2024-03-31 (report) |
| First duplicate in O(n) time and O(1) space | 1 | Reports: 2024-03-31 (report) |
| Inorder, preorder and postorder traversals | 1 | Reports: 2024-03-31 (report) |
| Search a sorted n×n matrix (sort rule not shared) | 1 | Reports: 2024-03-31 (report) |
| Check whether all nodes are at the same level (wording ambiguous) | 1 | Reports: 2024-03-31 (report) |
| Check whether two strings are rotations | 1 | Reports: 2024-03-31 (report) |
| Find the intersection node of two linked lists | 1 | Reports: 2024-03-31 (report) |
| Allocate and release a dynamic matrix | 1 | Reports: 2024-03-31 (report) |
| Multiply two matrices | 1 | Reports: 2024-03-31 (report); Additional same-company practice history: 2025-10-05 (report date), 2021-04-07, 2021-04-06 |
| Merge trees by summing corresponding nodes | 1 | Reports: 2024-03-31 (report) |
| Mirror a binary tree | 1 | Reports: 2024-03-31 (report) |
| Sum child-node values into a parent (scope not shared) | 1 | Reports: 2024-03-31 (report) |
Data Structures / Complexity Discussion
| Question | Frequency | All last seen dates |
|---|---|---|
| Restore fast search in a BST after many insertions | 1 | 2024-03-31 (report) |
07: SDE2: four rounds of trees, matrices and linked lists
Report date: March 31, 2024 · SDE2 · 2.6 years of experience · Offer
Round 1 — 60 minutes
The candidate implemented a queue using stack(s); the source says “a stack” without fixing the count. The practice exercise explicitly allows two stacks. AVL trees and their uses came up.
A custom BST had become slow to search after a large stream of insertions. The candidate proposed balancing it with AVL or red-black-tree techniques.
Round 2 — 45 minutes
Trapping Rain Water, followed by checking whether a binary tree is height-balanced.
Round 3 — 45 minutes
Find the first duplicate in an array in O(n) time and O(1) extra space. The account does not give the value range or permission to modify the input. The practice contract declares values 1..n, the earliest second occurrence, and temporary input mutation with restoration.
The candidate also implemented inorder, preorder and postorder traversals, then counted ways to climb stairs using steps of size one or two.
Round 4 — about two hours
The tasks were:
- Search for a value in a sorted n×n matrix; the sorting convention was not given. The linked exercise explicitly uses sorted rows and columns and permits rectangular inputs.
- Check whether “all nodes” in a tree are at the same level. The linked exercise takes that wording literally, including the root and internal nodes; the source does not clarify whether it meant leaves.
- Check whether two strings are rotations.
- Find the intersection node of two linked lists. The practice encoding preserves shared node identity.
- Allocate and release a dynamically sized matrix. The practice task specifies allocation, read/write and explicit release operations.
- Multiply two matrices.
- Merge two trees by summing corresponding nodes.
- Create a mirror image of a binary tree.
- Sum child-node values into their parent; the source does not distinguish immediate children from all descendants. The linked variant explicitly uses all strict descendants and excludes the node’s own value.
Senior Software Engineer - AI
1 account in this role.
What came up
Reported topics: coding tasks; projects or prior work; GPU or ML questions; C/C++ explicitly reported.
Questions we found
Behavioral (BQ)
| Question | Frequency | All last seen dates |
|---|---|---|
| Location and career/role roadmap | 1 | 2024 (interview year) |
ML Questions
| Question | Frequency | All last seen dates |
|---|---|---|
| ReLU versus PReLU | 1 | 2024 (interview year) |
| Regularization | 1 | 2024 (interview year) |
| Object-detection workflow | 1 | 2024 (interview year) |
| Residual connections | 1 | 2024 (interview year) |
| Non-max suppression | 1 | 2024 (interview year) |
| Transformer architecture | 1 | 2024 (interview year) |
| Attention types and applications | 1 | 2024 (interview year) |
08: Senior AI engineer: six rounds, with a long gap before the callback
Interviews: 2024 · Report date: December 21, 2024 · Senior Software Engineer, AI · Hyderabad role; candidate preferred Bengaluru
Round 1 — 60 minutes
Projects, challenges and implemented solutions, followed by a graph-based problem whose operation was not shared.
Round 2 — ML, 60 minutes
ReLU/PReLU, regularization, an object-detection workflow, residual connections and non-max suppression.
The candidate had joined Intel during the gap before AMD called back.
Round 3 — 45 minutes, after a callback about 2.5 months later
An easy DSA question was coded in C++, and smart pointers were discussed. The exact coding question was not named.
Round 4 — ML, 60 minutes
Transformer architecture, types of attention and their applications.
Round 5 — manager, 45 minutes
ML contributions, technical skills and team fit.
Round 6 — HR, 45 minutes
Location and the career/role roadmap. The candidate's decision was still open in the account.
MTS Software Development Engineer - GPU
1 account in this role.
What came up
Reported topics: coding tasks; projects or prior work; OS, memory, architecture or concurrency; GPU or ML questions.
09: Taipei GPU software: a two-hour hiring-manager deep dive
Interview period: March–June 2023 job search · MTS Software Development Engineer · Nangang/Taipei · About 3 years of prior GPU-software experience
The hiring-manager interview lasted about two hours. The candidate's prior GPU/OpenCL work and each resume experience received detailed questions.
Graphics kernels led into assembly-level optimization and computer architecture. There was a bit-manipulation coding question and a matrix coding question, with optimization expected for both, but the exact operations were not shared.
There was no final decision in the account. The MTS title alone does not establish a universal staff-level equivalent.
Role not reported — post-offer start delay
1 account in this role.
What came up
This account describes a start-date concern after an offer, while export-license and H-1B transfer paperwork were pending.
10: After the offer: export-license and transfer paperwork delayed the start
Report: Mar 8, 2023 · Role not reported · Previously employed; laid off in January · AMD offer received
The candidate had been laid off in January and had recently received an AMD offer. Export-license processing and an H-1B transfer were underway. Their previous employer's payroll would continue until mid-March, and they were worried about the effect of a delayed start on their own immigration timeline.
This is a post-offer account. It contains no technical interview questions, round descriptions or final resolution of the paperwork. The questions in the post were the candidate's requests for immigration advice, rather than questions asked by AMD interviewers.
Software System Designer 2 — Test Automation
1 account in this role.
What came up
One test-automation candidate received three named tasks across two virtual sessions: Two Sum, a SQL join and LeetCode 3.
Questions we found
Algorithms / Data Structures
| Question | Frequency | All last seen dates |
|---|---|---|
| Two Sum | 1 report | Reports: 2022-06-14 (report); Jan–Mar 2022 (header interview period); Additional same-company practice history: 2025-01-06 |
| Longest substring without repeating characters | 1 report | Reports: 2022-06-14 (report); Jan–Mar 2022 (header interview period) |
SQL / Data Questions
| Question | Frequency | All last seen dates |
|---|---|---|
| Write an inner join; later SQL parts were not disclosed | 1 report | Reports: 2022-06-14 (report); Jan–Mar 2022 (header interview period) |
11: Markham test automation: Two Sum, part of a SQL exercise, then LeetCode 3
Interview period: Jan–Mar 2022 (header) · Report: Jun 14, 2022 · Software System Designer 2, Test Automation · Markham, Canada · Offer
This was a cloud-team Test Automation role in Markham. A 30-minute recruiter phone screen covered the candidate's background.
First virtual technical session
The manager talked about the candidate and the team's work. Two senior engineers then discussed previous projects. Roughly the final 40 minutes were spent in CoderPad on Two Sum and a SQL question with several parts.
The candidate completed only the inner-join part. They expected the unfinished SQL exercise to end the process, but were told a few days later that they had passed. The later SQL prompts were not included in the account.
Next virtual session
A senior engineer asked Longest Substring Without Repeating Characters, identified in the report as LeetCode 3. The rest was a short conversation. An offer followed after the weekend.
Verification — PCIe
1 account in this role.
What came up
Two technical rounds concentrated on verification experience and PCIe. The author dates the move to 2022.
Questions we found
Verification / PCIe Questions
| Question | Frequency | All last seen dates |
|---|---|---|
| How would you build a testbench? | 1 report | 2026-10-01 (report); 2022 (author’s interview year; conflicting header is not a second sighting) |
| What are PCIe TS1 and TS2 used for? | 1 report | 2026-10-01 (report); 2022 (author’s interview year; conflicting header is not a second sighting) |
| When can different PCIe transactions be reordered? | 1 report | 2026-10-01 (report); 2022 (author’s interview year; conflicting header is not a second sighting) |
12: Vancouver verification: a remote loop with a deep PCIe round
Interview year: 2022, according to the author · Report: Oct 1, 2026 · Verification · Vancouver, Canada · Joined AMD
The author said they moved from another chip company to AMD in 2022. The structured header instead lists Apr–Jun 2024; this account keeps the author's stated year and does not turn the October 2026 publication into a new interview.
There were three video interviews, including two technical rounds of about an hour each. The author connected the video format to the Vancouver office arrangements at the time.
Verification questions covered SystemVerilog and UVM basics, verification methodology, building a testbench, scoreboards, monitors, coverage and RTL fundamentals. Interviewers went into the candidate's own projects in detail.
A manager based in Toronto handled a deeper PCIe discussion: TS1/TS2, LTSSM and link training, ordering, posted and non-posted requests, completions, TLPs and transactions. One concrete follow-up was when transactions could be reordered.
The report does not disclose an executable coding task or a complete protocol-design prompt.
Role not reported — C and memory panel
1 account in this role.
What came up
An unexpectedly large panel combined reverse-bit coding with C layout and operating-system questions.
Questions we found
Algorithms / Data Structures
| Question | Frequency | All last seen dates |
|---|---|---|
| Reverse bits | 1 report | Reports: 2020-05-20 (report); Jan–Mar 2020 (header interview period); Additional same-company practice history: 2021-04-06 |
C / Operating-System Questions
| Question | Frequency | All last seen dates |
|---|---|---|
| Compare big-endian and little-endian representation | 1 report | 2020-05-20 (report); Jan–Mar 2020 (header interview period) |
| Compare threads with processes | 1 report | 2020-05-20 (report); Jan–Mar 2020 (header interview period) |
| Calculate the size of a supplied structure and a union | 1 report | 2020-05-20 (report); Jan–Mar 2020 (header interview period) |
| Describe stack, heap and BSS layout | 1 report | 2020-05-20 (report); Jan–Mar 2020 (header interview period) |
Behavioral (BQ)
| Question | Frequency | All last seen dates |
|---|---|---|
| Introduce yourself | 1 report | 2020-05-20 (report); Jan–Mar 2020 (header interview period) |
13: A Skype panel expected to have three interviewers opened with eight
Interview period: Jan–Mar 2020 (header) · Report: May 20, 2020 · Role not reported · Experienced-hire context · Rejected
The recruiter had described a two-hour Skype interview with three interviewers. Eight members of the team appeared at the start of the session.
After the introduction, the candidate was asked about bit manipulation and reversing bits. Other questions compared big- and little-endian representation, threads and processes, and the layout of stack, heap and BSS memory.
The panel supplied a structure and asked for its memory size, then asked about a union. The declarations were not retained, so the report cannot establish the particular padding or alignment calculation. SHA and RSA key-related questions also came up, without enough wording to reconstruct them.
The structured record marks this application as unsuccessful. The exact job title was not given.
Comments
0No comments yet — be the first to share what you know.