Snap Machine Learning Engineer interview questions

39 Machine Learning Engineers have shared notes from their Snap interviews, with the questions they were asked in each round. That is 33% of the 119 Snap interview notes on Screna AI.

39 Machine Learning Engineer interview notes

Snap·Machine Learning Engineer·Onsite - System Design / Architecture

Jun 2026
System design round for an MLE role, centered entirely on one massive open-ended problem about building a product tagging pipeline from scratch. Felt more like a staff-level architecture session than a typical ML interview. Walked away unsure if I covered enough ground or just rambled for an hour.
  • Design an end-to-end pipeline that automatically assigns category, attribute, and quality tags to products listed on a short-video commerce platform, starting from raw seller inputs like titles, descriptions, images, and video.
  • How would you design the taxonomy and label space for product tags, and how do you decide between training separate models per tag type versus a single joint model?
  • Walk through how you'd handle weak supervision and human-in-the-loop labeling when you don't have enough ground truth data to train a reliable tagger.
  • How do you monitor tag quality in production, and what signals would you track to detect model drift or degradation over time?
  • How do the product tags you generate get consumed by downstream systems like search, recommendations, and ads, and does that change how you design the tagging pipeline?
  • How would you handle multilingual product catalogs and adversarial sellers who try to manipulate tags to game search or recommendations?

“This was the whole interview.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Jun 2026
Interviewed for an MLE role at Snapchat and got hit pretty hard on normalization fundamentals. Not a bad experience overall but I clearly needed to sharpen up on the theory side before walking in.
  • Walk me through how Layer Normalization works, including how statistics are computed, what the learnable parameters do, and how it compares to Batch Normalization. Why do Transformers use LayerNorm instead of BatchNorm?

“I started okay on the mechanics: mean and variance computed per sample across the feature dimension, then you normalize and apply the learned scale and shift.”

View Post

Snap·Machine Learning Engineer·Onsite - System Design / Architecture

Jun 2026
Snapchat ML Engineer interview focused entirely on designing a recommendation system for Lenses. The depth they expected was real, covering everything from candidate generation to serving pipelines, and the follow-ups on cold start and multi-signal objectives were where things got genuinely hard.
  • Design a recommendation system for Snapchat Lenses, covering problem framing, candidate generation, ranking model architecture, feature design, label/objective definition, and training and serving pipelines.
  • How would you handle cold start for brand-new lenses with no engagement history, and separately for new users with no behavioral history?
  • Users interact with lenses at different levels: viewing, trying, sharing, saving. How would you model these multiple engagement signals? Walk through the trade-offs between a single model with weighted loss terms versus separate models per action combined into a final score.

“This is a big open-ended question and I spent probably too long on problem framing before they nudged me toward the actual system.”

View Post

Snap·Machine Learning Engineer·Onsite - System Design / Architecture

Jun 2026
Snapchat ML engineer interview that went deep on multimodal retrieval for short-video recommendations. The whole session was basically one long system design question with a lot of sub-parts, and I felt like I was treading water by the end of it.
  • Design a short-video recommendation and retrieval system using ML, given that only 20% of videos have text descriptions.
  • Walk through the query-side retrieval pipeline end to end for this system.
  • How do CLIP-style joint vision-text embeddings work, and what loss functions are used in contrastive learning for training them?
  • What are the main disadvantages of pure embedding-based retrieval, and how would you address them?
  • How would you mitigate popularity bias in this recommendation system?

“This is the kind of question where the constraint is the whole point and I almost glossed over it.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Jun 2026
Snapchat ML engineer interview, coding round focused on a geometry problem that turned into a pretty detailed algorithms discussion. The interviewer pushed hard on complexity tradeoffs which I wasn't fully prepared for.
  • Given an array of 2D points and an integer k, return the k points closest to the origin. Walk through an O(N log K) heap-based solution and explain why it's preferable to sorting when k is much smaller than n.

“I knew the heap approach going in but fumbled the explanation of why it beats sorting.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Jun 2026
ML fundamentals interview at Snapchat, pretty much a gauntlet of core concepts with the interviewer pushing deeper at every turn. Three main topics but each one had layers, so you couldn't just recite a definition and move on.
  • Explain the bias-variance trade-off. How do model capacity, training set size, and regularization each affect it, and what does it look like on a learning curve?
  • How do you handle class imbalance in a training dataset, and how do you decide which metrics to use when evaluating a model trained on imbalanced data?
  • Walk me through how you diagnose overfitting in a model and what steps you take to address it.

“Started fine on definitions but then they asked me to tie it to learning curves specifically and I kind of fumbled the connection between high-variance models and the gap between train vs validation loss.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Jun 2026
Snapchat ML Engineer interview that came down to a classic algorithmic problem. Nothing groundbreaking but the pressure of doing it live is a different beast than grinding it on your own.
  • Given an integer array and a target value, find all unique quadruplets in the array that sum to the target.

“I knew the two-pointer approach from doing 3Sum before so extending it to four numbers felt doable in theory.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Jun 2026
Snapchat ML Engineer interview with a classic recursion problem that feels deceptively easy until you're staring at a blank editor. The backtracking logic is standard but explaining the constraint clearly under pressure is a different thing.
  • Given an integer n, generate all combinations of n pairs of well-formed parentheses and return them as a list of strings.

“I knew this problem but still fumbled the explanation of why you can only place a closing paren when the close count is below the open count.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Jun 2026
Went through a technical screen for an MLE role at Snapchat and got hit with a fairly deep question on handling class imbalance. Not a casual conversation, they clearly wanted to see if you knew the full toolkit, not just SMOTE.
  • How do you handle imbalanced datasets in a classification problem? Walk through your approach.

“I started with resampling, talked about oversampling the minority class and undersampling the majority, mentioned synthetic generation techniques.”

View Post

Snap·Machine Learning Engineer·Technical Phone Screen

Jun 2026
Snapchat ML engineer interview that leaned pretty hard into the theoretical side of probabilistic modeling. One question on calibration took up a big chunk of the conversation and went deeper than I expected.
  • What is model calibration, how do you measure it, and what can you do to improve it?

“This one sprawled out way more than I anticipated.”

View Post