
Snap Machine Learning Engineer Interview Questions
& Process
Real candidates share what happened, how many rounds they had,
and how the experience turned out.
Based on 31 interview experiences · FREE TO READ
Candidate interview experiences
First-hand accounts from people who interviewed at Snap.
ML Engineer
The discussions on Deep Generative AI and the technical nuances of the topics were very engaging. The coding problems were challenging and thought-provoking. Also, the conversation on the ML Systems Designing in production stage was particularly insightful for real-world applications.
- Can you explain how to utilize multiple GPUs for training and discuss optimization techniques for deep learning inference?
Machine Learning Engineer
I applied online and heard back from a Snap recruiter about two weeks later for an initial screen. The technical screen lasted 60 mins: 10 mins intro about my background and projects, 10 mins on ML fundamentals, 10 mins behavioral, and 30 mins for a LeetCode medium problem which the interviewer later changed to hard. The onsite interview was planned with the following structure: ML fundamentals (60 mins), ML Applied (60 mins), Informal Q&A with MLE (30 mins), ML System Design (60 mins), Coding (60 mins), and another Coding round (60 mins). The interviewers stuck to a fixed set of questions and weren't interested in my projects or discussing models.
- Tell me about basic ML questions like metrics, unbalanced data, overfitting, and optimizers.
- What are the trade-offs between different modeling approaches for a real-world problem?
- How would you design recommender systems in production?
Senior Machine Learning Engineer
So I had one phone call first. Then, the onsite was with six interviewers total. Two of them were coding interviews, which were pretty difficult, like LC hard level. Then there was one on ML fundamentals, one on ML design, and one on ML system design. Lastly, there was a behavioral interview. Oh, and I met one of the coding interviewers, but his English was really hard to follow, and the problem he gave me, running median, was kind of boring. If I hadn't seen it before, I doubt I could have come up with the best solution.
- How would you implement a running median algorithm?
- Can you solve the Circle in a matrix problem?
Snap Machine Learning Engineer Interview Questions
Quoted word for word from Snap interview reports.
“Can you explain the loss function for GANs?”
Read reports →“Can you implement a convolutional operator, for instance, using a for loop in Python?”
Read reports →“Can you explain how to implement np.sum() along a specified axis?”
Read reports →“Describe the design of a neural network capable of determining the maximum of two input numbers.”
Read reports →“Can you write the code for a CNN forward pass?”
Read report →“Can you explain how to create a random 3D rotation matrix?”
Read report →“Can you write code to blur an image?”
Read report →“Could you write the code for a convolution neural network?”
Read report →“Can you tell me about the time and memory efficiency of pandas join operations?”
Read report →Formats, difficulty and experience
Across all 31 Snap interview reports.