AMD · Onsite
In processAMD Intern / Co-op OA / Interviews by Role through Oct 2026: 18 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: DevOps Engineer; Systems Engineer; Embedded Firmware Engineer; Software Engineer; GPU / Compiler Software Intern; Video Encode/Decode Software Engineer.
- Part 2 (current): ML Software Engineer; Software Custom Engineering; Internship role not specified - MTech; Compiler Engineer; GPU Architecture; Virtualization Validation; Internship role not specified - dual-degree campus.
Coverage: 7 of 18 AMD Intern / Co-op accounts in 7 separate role sections.
Other AMD hiring levels
- Experienced Full-Time: 16 accounts in 3 parts.
- Hiring Level Not Reported: 9 accounts.
- New Grad: 11 accounts in 2 parts.
Historical Xilinx reports
Intern / Co-op at Xilinx covers 2 accounts under their original employer, with separate counts.
Roles in this article
- ML Software Engineer — 1 account.
- Software Custom Engineering — 1 account.
- Internship role not specified - MTech — 1 account.
- Compiler Engineer — 1 account.
- GPU Architecture — 1 account.
- Virtualization Validation — 1 account.
- Internship role not specified - dual-degree campus — 1 account.
ML Software Engineer
1 account in this role.
What came up
Reported topics: coding tasks; OS, memory, architecture or concurrency; projects or resume; GPU, ML or hardware-specific topics.
Questions we found
Algorithms / Data Structures
| Question | Frequency | All last seen dates |
|---|---|---|
| Matrix multiplication in C++ | 1 | Reports: 2025-11 (interview); Additional same-company practice history: 2025-10-05 (report date), 2024-03-31, 2021-04-07, 2021-04-06 |
12: An ML software internship with C++ matrix multiplication
Interview: November 2025 · ML Software Engineer Intern · Markham, Canada · Selected
In about an hour, the candidate explained project design choices and possible improvements, discussed OS and GPU threads, and implemented matrix multiplication in C++. The report does not give dimensions, bounds or additional matrix rules. The practice exercise defines those bounds and adds explicit parallel variants.
Software Custom Engineering
1 account in this role.
13: A 2024 Taiwan software-custom-engineering internship timeline
Interview year: 2024 · Software Custom Engineering Intern · Taiwan · Rejected
The engineer interview took place at D+6 on a software-driver/ML-acceleration team. The candidate heard they had passed at D+9, met the manager for a behavioral round at D+12, and received a rejection at D+28.
No exact technical question was given. The AMD entry is in the author's 2024 section, although the wider page also covers later years.
Internship role not specified - MTech
1 account in this role.
What came up
Reported topics: language, runtime or compiler details; coding tasks; OS, memory, architecture or concurrency; projects or resume; GPU, ML or hardware-specific topics.
Questions we found
Algorithms / Data Structures
| Question | Frequency | All last seen dates |
|---|---|---|
| Recursive factorial | 1 | Reports: 2021-06-29 (same account’s report publication); 2021-04 (interview) |
| Deduplicate MALAYALAM to MALY, then avoid a dictionary | 1 | Reports: 2021-06-29 (same account’s report publication); 2021-04 (interview) |
| List comprehension for even-indexed elements | 1 | Reports: 2021-06-29 (same account’s report publication); 2021-04 (interview) |
| Square a list, then use a dictionary | 1 | Reports: 2021-06-29 (same account’s report publication); 2021-04 (interview) |
| Print 1–100 in C without loops | 1 | Reports: 2021-06-29 (same account’s report publication); 2021-04 (interview) |
| Use static in a circle-radius program (input not shared) | 1 | Reports: 2021-06-29 (same account’s report publication); 2021-04 (interview) |
C / Compiler Diagnostics
| Question | Frequency | All last seen dates |
|---|---|---|
| Find C allocation/compile mistakes | 1 | 2021-04 (interview) |
Behavioral (BQ)
| Question | Frequency | All last seen dates |
|---|---|---|
| Introduction and strongest/preferred programming language | 1 | 2021-04 (interview) |
C/C++ / OS Questions
| Question | Frequency | All last seen dates |
|---|---|---|
| Linux commands for RAM, file sizes and CPU use | 1 | 2021-04 (interview) |
| Find a C type's size with and without sizeof | 1 | 2021-04 (interview) |
ML Questions
| Question | Frequency | All last seen dates |
|---|---|---|
| Max pooling and ordering fully connected layers | 1 | 2021-04 (interview) |
| Activation functions: sigmoid versus ReLU | 1 | 2021-04 (interview) |
14: MTech internship: Python expressions, C mistakes and CNN questions
Interview: April 2021 · Nine-month MTech internship · CS/ML background · Not selected
Two interviewers conducted roughly two hours of interviews, with a five-minute break.
First interviewer
After an introduction and strongest-language question, they discussed a CNN project: its layers, conclusions and future work, max pooling, the order of fully connected layers, activation functions, and sigmoid versus ReLU.
A C snippet used an integer ptr for malloc(n*sizeof(int)), lacked a semicolon on that line, printed with %d, and included only stdio. The candidate had to find the compile/allocation mistakes.
Python tasks included taking the elements at even indices of [2,5,3,4,6,9,10] using a list comprehension, then writing recursive factorial.
Second interviewer
The introduction and preferred-language question returned, followed by project inferences and future work.
They were asked for Linux commands to inspect RAM, file sizes and CPU usage, and how to determine a C type's size with and without sizeof.
Further tasks were to map [1,2,3,4,5] to [1,4,9,16,25] and then use a dictionary; print 1 through 100 in C without loops; and remove repeated letters from MALAYALAM to obtain MALY, with a follow-up avoiding a dictionary.
The linked sequence exercise uses a portable function that returns 1 through a supplied bound, with loop-free references. The original request was to print 1 through 100 in C.
C storage classes came up. A final task involved using static in a C program to find a circle’s radius, but the source does not give the input or inverse relation. The linked exercise explicitly assumes area as input. Its C++ references use a function-local static PI constant; the other languages use constant adapters, and numeric grading does not check C syntax.
The author reported that 23 of 40 interviewed applicants were selected from an initial pool of about 750, and that they were not among those selected.
Compiler Engineer
1 account in this role.
What came up
Reported topics: language, runtime or compiler details; OS, memory, architecture or concurrency; projects or resume; tooling, CI/CD or infrastructure.
Questions we found
C/C++ / OS Questions
| Question | Frequency | All last seen dates |
|---|---|---|
| Debugging with LLDB; candidate discussed GDB instead | 1 | Post displayed as 5 years old on 2026-10-30; exact date unavailable |
15: Compiler co-op: Linux, LLVM and a switch from LLDB to GDB
Post displayed as 5 years old on October 30, 2026; exact date unavailable · Compiler Engineer co-op/intern · India · Joined
The candidate explicitly said they were not asked generic DSA questions.
Instead, the interviews covered GNU/Linux, their Clang/LLVM work with Arduino, GCC/GLibC work, compiler flags and shell scripting. Asked about debugging with LLDB, they explained GDB because they were unfamiliar with LLDB.
GPU Architecture
1 account in this role.
What came up
Graphics APIs and GPU architecture drove this internship screen. Shader stages and spatial hierarchies were also discussed.
Questions we found
Graphics / Architecture Questions
| Question | Frequency | All last seen dates |
|---|---|---|
| Discuss DirectX 12 and Vulkan features compared with earlier APIs | 1 report | 2018-11-08 (report; header period conflicts) |
| Describe GPU components, the graphics pipeline and cases where a CPU can be faster | 1 report | 2018-11-08 (report; header period conflicts) |
| Explain the inputs, outputs and work of vertex and pixel shaders | 1 report | 2018-11-08 (report; header period conflicts) |
| Compare KD trees with a bounding-box hierarchy and discuss when the hierarchy has an advantage | 1 report | 2018-11-08 (report; header period conflicts) |
16: GPU Architecture internship: graphics APIs, shaders and spatial structures
Report: Nov 8, 2018 · GPU Architecture internship · Hiring outcome not stated
The author explicitly called this an internship interview with the GPU Architecture team. They had a graphics background and had been moving toward software roles, but said they were not yet clear on the team's work.
The interview asked about DirectX 12 and Vulkan and their newer features. It moved to GPU components, the GPU computation pipeline, why GPUs can be fast, and conditions under which a CPU could be faster.
Graphics questions covered what vertex and pixel shaders receive, produce and do. The candidate was also asked about KD trees and a bounding-box hierarchy, including when the latter has an advantage.
The body describes the interview as happening on the report day in November 2018, while the structured header lists Jul–Sep 2019. The dates conflict; no later interview is inferred from the header. The report does not name a coding exercise.
Virtualization Validation
1 account in this role.
What came up
Reported topics: language, runtime or compiler details; OS, memory, architecture or concurrency; tooling, CI/CD or infrastructure.
17: Virtualization co-op: a five-person second panel
Account available by March 28, 2025; exact interview date unknown · Virtualization validation co-op · Markham, Canada · Selected
The candidate described two interviews. Five interviewers attended the second panel, which covered virtualization, Python, system architecture, server management, Docker and hypervisors.
They reported receiving an offer about 30 minutes later. No executable task was shared, and the guide's later update is not another interview date.
Internship role not specified - dual-degree campus
1 account in this role.
What came up
Reported topics: language, runtime or compiler details; coding tasks; OS, memory, architecture or concurrency; projects or resume.
Questions we found
Algorithms / Data Structures
| Question | Frequency | All last seen dates |
|---|---|---|
| Insert a node efficiently into the middle of a linked list | 1 | Reports: Date not reported; page last updated 2021-03-09 is only an availability bound |
C/C++ / OS Questions
| Question | Frequency | All last seen dates |
|---|---|---|
| Where is a pointer stored? | 1 | Date not reported; page last updated 2021-03-09 is only an availability bound |
| Threads versus processes | 1 | Date not reported; page last updated 2021-03-09 is only an availability bound |
| Virtual versus physical address of the pointer in the example | 1 | Date not reported; page last updated 2021-03-09 is only an availability bound |
| Boot sequence and bootloader duties | 1 | Date not reported; page last updated 2021-03-09 is only an availability bound |
18: A nine-month internship: middle insertion, pointers and address translation
Page last updated March 9, 2021; exact interview date unknown · Nine-month internship · Dual-degree CS campus applicant
On Microsoft Teams, the candidate first explained a past project, then had to insert a node efficiently in the middle of a linked list. The linked practice contract specifies which middle position to use for odd and even lengths.
The coding discussion expanded into where the pointer itself is stored and the virtual versus physical address of that pointer. Further questions covered threads versus processes, synchronization of shared resources, virtual-memory management, virtual-address translation, the boot sequence, a bootloader's duties and why caches exist.
In this series
- Part 1
- Part 2 (current)
Comments
0No comments yet — be the first to share what you know.