Google Machine Learning Engineer interview questions

97 Machine Learning Engineers have shared notes from their Google interviews, with the questions they were asked in each round. That is 5% of the 2,146 Google interview notes on Screna AI.

97 Machine Learning Engineer interview notes

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

Jul 2026
Machine learning system design round at Google for an MLE role. The whole thing was one big open-ended design question about building a real-time recommendation system, and they expected you to drive the conversation across a lot of different dimensions.
  • Design a real-time recommendation system for a large-scale mobile app that serves personalized results to millions of daily active users with sub-100ms P95 latency.
  • How would you design the feature store to ensure consistency between offline training and online inference?
  • Walk me through how you'd handle cold start for both new users and new items.
  • Given a strict latency budget, where would you use caching in the serving path and what TTL and consistency strategies would you apply?
  • What offline and online metrics would you track, and how would you set up experiments to evaluate the recommender?
  • How would you design the system to degrade gracefully if the model server is slow or unavailable?
  • Propose a baseline model and then a more advanced ranking model, and explain why you'd structure it as candidate generation followed by ranking.

“This was the entire interview, basically.”

View Post

Google·Machine Learning Engineer·Onsite - Coding / Algorithms

Jul 2026
Google ML Engineer interview, technical round focused entirely on implementing a Transformer block from scratch. Pretty deep dive, they wanted working code plus a real understanding of the math behind it.
  • Implement a multi-head self-attention module in Python (NumPy or PyTorch), taking an input of shape (batch_size, sequence_length, hidden_dim), splitting into multiple heads, computing scaled dot-product attention with optional masking, and projecting the concatenated output.
  • Implement a SwiGLU feed-forward layer where the gate and value projections run in parallel, apply Swish activation to the gate path, multiply elementwise, then project the result.
  • Put the attention and feed-forward components together into a full Transformer block with residual connections and layer normalization, returning output with the same shape as the input.
  • Explain the time and memory complexity of multi-head attention, and walk through how masking works across the attention heads.

“The shape tracking is where things get slippery.”

View Post

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

Jul 2026
Big ML system design question at Google for an MLE role. The whole interview was basically one giant open-ended problem about predicting a continuous target from chemical reaction data, and they expected you to go deep on every layer from EDA all the way to production monitoring. Felt more like a whiteboard dissertation than a typical interview.
  • You have a dataset of chemical reactions with molecule names and a real-valued reaction factor. Walk through your full end-to-end ML approach: from EDA and feature engineering to modeling, evaluation, and production deployment.
  • How would you split this chemical reaction dataset to properly evaluate generalization, and what are the tradeoffs between different splitting strategies?
  • How would you handle unseen molecules at inference time, and how does symmetry (A+B being equivalent to B+A) affect your modeling choices?
  • What modeling approaches would you consider for this task, from simple baselines up to deep learning, and how would you handle uncertainty estimation?
  • How would you design the production pipeline for this system, including monitoring, retraining triggers, and safety considerations given the scientific domain?

“This was the whole interview basically.”

View Post

Google·Machine Learning Engineer·Technical Phone Screen

Jul 2026
Went through a technical screen for an ML Engineer role at Google that was basically a deep dive into how LLMs actually work, soup to nuts. One long question that branched into a dozen sub-topics. Felt like a PhD oral exam more than an interview.
  • Walk through the full lifecycle of a large language model, from data collection through deployment, covering pretraining, architecture choices, fine-tuning, and preference optimization. What are the key trade-offs across quality, safety, latency, cost, and scalability?

“This is the kind of question that sounds manageable until you're three minutes in and realize you've only covered data filtering and you have like eight more topics to go.”

View Post

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

Jul 2026
Google ML Engineer system design round focused entirely on building a recommendation system from scratch, covering pretty much every angle you can imagine. It was one of the more exhaustive design questions I've sat through, the kind where you realize halfway in that you've been talking for 20 minutes and still haven't touched ranking.
  • Design a recommendation system for a large consumer platform that supports both e-commerce product feeds and video media feeds.

“This is a monster of a question.”

View Post

Google·Machine Learning Engineer·Technical Phone Screen

Jul 2026
Interviewed for an ML engineer role at Google and got a string manipulation problem that looked deceptively simple at first glance.
  • Given two strings of equal length, is it possible to cut each string in half and combine the halves to form a palindrome?

“Spent the first few minutes just staring at the constraints.”

View Post

Google·Machine Learning Engineer·Onsite - Coding / Algorithms

Jul 2026
Google ML Engineer interview with a coding round that went deeper than I expected. The string search problem looked familiar at first glance but they wanted more than a brute force pass.
  • Implement substring search: given a haystack and a needle, return the index of the first occurrence of the needle in the haystack, or -1 if it's not there. Naive O(n*m) won't cut it here.

“Started with the brute force and they let me finish before asking what the complexity was.”

View Post

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

Jul 2026
Google ML system design round focused entirely on building an app recommendation system for an app store. Pretty broad scope, they wanted you to go deep on basically every layer from retrieval to serving to experimentation.
  • Design an ML-powered app recommendation system for an app store, covering candidate generation, ranking, serving infrastructure, metrics, cold-start handling, and A/B testing.

“This is the whole interview in one question basically.”

View Post

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

Jul 2026
Did a system design round at Google for an MLE role. The question was around building a recommendation system and my feedback came back pretty lukewarm, so make of that what you will.
  • Design a recommendation system for products or videos.

“Felt okay in the moment but the feedback said otherwise.”

View Post

Google·Machine Learning Engineer·Onsite - Coding / Algorithms

Jun 2026
Google ML Engineer interview, got a coding round with a data structures problem that had a twist I didn't fully see coming. The LRU variant with pinning caught me off guard in the implementation details even if the concept felt familiar.
  • Design and implement a modified LRU cache that supports get, put, pin, and unpin operations, where pinned keys cannot be evicted until explicitly unpinned. All operations should run in average O(1) time.

“I knew LRU cold going in, doubly linked list plus a hashmap, done.”

View Post