IBM — Software Engineer ✅ Passed

Level: Senior-Level

Round: Online Assessment · Type: Coding · Difficulty: 6/10 · Duration: 60 min · Interviewer: Neutral

Topics: Convolutional Neural Networks, Multithreading, Optimization, Hardware Optimization

Location: San Francisco Bay Area

Interview date: 2025-05-15

Summary

I had a one-hour online assessment consisting of a coding question and a follow-up on multithreading optimization.

The position I was interviewed with: RedHat MLE, AI inference system

Details

After a brief self-introduction, I was given the following coding problem:

`

The problem we'd like you to work on is a "valid" 1-d convolution function.

That is, only perform the convolution on complete input regions.

Example:

input = 11, 2, 3, 4, 51 kernel = [2, 1, 0.51 bias = 0.5 output [0] = (1 * 2) + (2* 1) +13 * 0.5) +0.5=6 output [11 = (2 * 2) + (3 * 1) + (4 * 0.5) + 0.5 = 9.5 output [21 = (3 * 2) + (4 * 1) + (5 * 0.5) + 0.5 = 13 output = [6, 9.5, 13] `

The follow-up question involved multithreading optimization. I was asked about optimization methods, considering both hardware and CPU. I initially suggested NumPy vector slicing, but the interviewer prompted me to think about hardware. The expected answer was multithreading. I was then asked how to code for these three cases:

• Input size: 1 million, kernel size: 3
• Input size: 1 million, kernel size: 1 million
• Maximum thread count: 100

The assessment concluded with a general discussion about my experience contributing to open-source projects.