practice a little every day, paws crossed ♡
Machine Learning Engineer Interview Questions
Practice 9403 coding, system design and machine learning questions from 546 companies, sorted by role, topic and interview round. Every question has step-by-step hints, a code runner that supports 8 languages and an AI mentor; members also get the full solution walkthrough.
- 9403
- Questions
- 546
- Companies
- 11
- Topics
- 8
- Languages
Showing 91–120 of 202 results
- NetflixMediumMachine Learning EngineerExplain self-attention, LoRA, Adam vs SGD, ViT
This question evaluates understanding of modern Machine Learning/Deep Learning topics, including self-attention mechanics (queries, keys, values and scaled logits), Low-Rank Adaptation (LoRA) for parameter-efficient fine-tuning and memory savings, optimizer behavior (Adam versus SGD with momentum), and architectural trade-offs between Vision Transformers and CNNs including patch-size considerations. It is categorized under Machine Learning and is commonly asked because it probes both conceptual understanding and practical application—testing reasoning about training dynamics, model scaling, fine-tuning strategies, and resource/performance trade-offs.
Machine LearningTechnical Screen - NetflixMediumMachine Learning EngineerExplain tokenization and Transformer variants
This question evaluates understanding of tokenization techniques and Transformer architecture, covering competencies in subword and SentencePiece-style tokenizers, Transformer block internals, and comparisons of modern architectural variants and trade-offs.
Machine LearningTechnical Screen - NetflixHardMachine Learning EngineerL1 vs L2 Normalization: Definitions, Geometry and When to Use Each
A machine learning fundamentals question asking how L1 normalization differs from L2 normalization of a vector. It tests precise definitions, what each normalized vector represents geometrically, when each is the right choice, and the ability to keep normalization distinct from L1 and L2 regularization.
Machine LearningTechnical Screen - NetflixEasyMachine Learning EngineerML Basics Check: Bayes' Rule, Forward-Pass Arithmetic, KNN and Transformer Concepts
A set of basic machine learning checks: a Bayes' rule posterior with a low base rate, forward-pass and parameter-count arithmetic for small neural networks, and true-or-false concept checks on k-nearest neighbors and Transformer attention. It tests quick, accurate reasoning on core ML fundamentals.
Machine LearningOnline Assessment - NewsBreakMediumMachine Learning EngineerDerive and implement calibration via temperature scaling
This interview question evaluates core ML concepts, assumptions, math intuition, training/evaluation trade-offs, and practical failure modes in a realistic interview setting. A strong answer for Derive and implement calibration via temperature scaling states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.
Machine LearningTechnical Screen - NewsBreakHardMachine Learning EngineerDesign approach for class imbalance
This interview question evaluates core ML concepts, assumptions, math intuition, training/evaluation trade-offs, and practical failure modes in a realistic interview setting. A strong answer for Design approach for class imbalance states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.
Machine LearningTechnical Screen - NextdoorMediumMachine Learning EngineerBuild a model using only pandas/numpy
This question evaluates proficiency in applied machine learning competencies including data preprocessing, categorical encoding, feature scaling, implementing baseline models and optimization using numpy/pandas, and selecting appropriate evaluation metrics.
Machine LearningTechnical Screen - OpenAIHardMachine Learning EngineerBuild and troubleshoot image classification and backprop
Build and troubleshoot image classification and backprop evaluates core ML concepts, assumptions, math intuition, training/evaluation trade-offs, and practical failure modes in a realistic interview setting. A strong answer states assumptions, handles edge cases, explains trade-offs, and shows how to validate the result clearly.
Machine LearningTechnical Screen - OpenAIHardMachine Learning EngineerCompute Matrix Prefix Products And Gradients
This question evaluates understanding of automatic differentiation for matrix operations, specifically forward and reverse-mode gradients for sequential, associative (but noncommutative) matrix products, together with algorithmic analysis of time and space complexity.
Machine LearningOnsite - OpenAIMediumMachine Learning EngineerCompute entropy and implement 1-NN
This question evaluates numerical stability and streaming computation for entropy estimation from logits, along with efficient vectorized linear-algebra implementation of 1-nearest-neighbor inference and its interpretation as a neural-network forward pass under different distance metrics.
Machine LearningOnsite - OpenAIMediumMachine Learning EngineerDebug MiniGPT and Backpropagate Matmul
This question evaluates proficiency in debugging PyTorch transformer implementations, reasoning about attention mechanics and causal masking, implementing key–value caching for autoregressive decoding, and constructing provably correct backward passes for matrix multiplications and custom autograd.Functions.
Machine LearningTechnical Screen - OpenAIMediumMachine Learning EngineerDebug Transformer and Add KV Cache
This problem tests your understanding of transformer architecture by requiring you to fix a broken attention mechanism and implement a Key-Value cache to optimize autoregressive inference.
Machine LearningOnsiteTransformersAttentionInference - OpenAIHardMachine Learning EngineerDebug a Small PyTorch Transformer with Four Bugs, Then Add a KV Cache
Debug a small PyTorch transformer language model that hides four bugs, some of which crash and some of which silently ruin training, then add a key-value cache so generation computes only the newest token each step. It tests causal masking, positional encoding, training-loop order and cached decoding.
Machine LearningOnsite - OpenAIMediumMachine Learning EngineerDebug a Transformer Implementation and Implement a KV Cache for Decoding
Practice ML coding question on Transformer language models: find and fix several deliberately planted bugs in a decoder-only Transformer implementation, then implement key-value caching from scratch for autoregressive generation. Tests attention mechanics, tensor-shape reasoning, causal masking, position handling, and verification against full recomputation.
Machine LearningTechnical Screen - OpenAIHardMachine Learning EngineerDebug a broken Transformer implementation
This question evaluates a candidate's ability to debug and validate a Transformer implementation, focusing on attention masking, parameter initialization, loss alignment, and other implementation-level correctness issues.
Machine LearningOnsite - OpenAIHardMachine Learning EngineerDebug a transformer training pipeline
An OpenAI ML engineer technical-screen debugging question: given a PyTorch Transformer training pipeline whose loss diverges and validation accuracy stays near chance (plus mask IndexErrors, AMP crashes, and nondeterminism), identify the distinct bugs and fix them. It tests Transformer internals - attention/causal masking, teacher-forcing label shift, positional encodings, ignore_index loss, optimizer and GradScaler order, determinism, and DDP - along with writing unit tests and a structured debugging plan.
Machine LearningTechnical Screen - OpenAIHardMachine Learning EngineerDebug transformer and train classifier
This question evaluates debugging and practical implementation skills for transformer-based text classification, covering model correctness, training loop integrity, data preprocessing, and evaluation metrics within the Machine Learning / Natural Language Processing domain.
Machine LearningTechnical Screen - OpenAIMediumMachine Learning EngineerDefend a Research Direction and Experiment Design
This question evaluates a candidate's ability to synthesize the state of the art in Machine Learning, defend a research direction, and design rigorous experiments, measuring competencies in literature analysis, methodological justification, experimental design, and technical communication.
Machine LearningOnsite - OpenAIHardMachine Learning EngineerDerive Backpropagation for Matrix-Product Layers
This question evaluates understanding of backpropagation through linear blocks, matrix calculus, and the computation of gradients with respect to individual weight matrices in a product of trainable matrices, including reasoning about tensor shapes and matrix differentials.
Machine LearningTechnical Screen - OpenAIMediumMachine Learning EngineerDerive Sharded Matrix Multiplication and Backpropagation
Derive forward and backward computation for column-sharded matrix multiplication across multiple devices. Track every local tensor shape, identify the collective needed for replicated input gradients, and propose tests for shard ordering, scaling, and numerical errors.
Machine LearningOnsite - OpenAIHardMachine Learning EngineerDiagnose Transformer training and inference bugs
This question evaluates practical debugging skills for Transformer-based sequence models, targeting machine learning engineers who must diagnose intermittent shape/dtype errors and convergence failures simultaneously. It tests deep familiarity with the full training stack — data pipelines, attention masking, positional encoding, optimizer precision, and distributed setup — and the ability to reason about how a single root cause can manifest across multiple symptoms.
Machine LearningTechnical Screen - OpenAIMediumMachine Learning EngineerFilter Bad Human Annotations
Tests the ability to detect and filter low-quality human labels in large-scale dataset pipelines by modeling annotator reliability, distinguishing genuine ambiguity from noise, and evaluating downstream model performance.
Machine LearningOnsitedata-centric-airlhflatent-variable-models - OpenAIHardMachine Learning EngineerHand-Rolled Backprop: Running Product and Stable Cross-Entropy
Derive and implement, without an autodiff library, the backward pass of a running (cumulative) product in linear time without division, then a numerically stable softmax cross-entropy and its gradient. Tests chain-rule fluency, prefix/suffix tricks, and gradient checking.
Machine LearningOnsiteautodiffbackpropagationnumerical-stability - OpenAIHardMachine Learning EngineerHand-Written Gradients: Prefix-Product Backward Pass and Stable Cross-Entropy
Write gradients by hand: the forward and backward pass of the prefix-product operation, in linear time and without division so inputs may contain zeros, and a numerically stable softmax cross-entropy loss with its gradient in NumPy. It tests the chain rule, prefix and suffix products, and gradient checking.
Machine LearningOnsite - OpenAIMediumMachine Learning EngineerImplement 1NN with NumPy
Tests your ability to implement a fast, vectorized 1-Nearest Neighbor (1NN) classifier using NumPy, focusing on memory-efficient distance computations and tie-breaking.
Machine LearningTechnical Screennumpyvectorizationmachine-learning - OpenAIMediumMachine Learning EngineerImplement Masked Cross-Entropy with Label Smoothing
Implement a numerically stable, masked cross-entropy loss function with label smoothing and temperature scaling in NumPy, returning both the loss and analytical gradients.
Machine LearningTechnical Screenmachine-learningnumpydeep-learning - OpenAIMediumMachine Learning EngineerImplement Prefix Products and Their Backward Pass
Work through inclusive prefix products in separate-output, in-place, reverse-gradient, and parallel variants. Explain zeros, overflow, mutation and aliasing, constant temporary workspace, synchronization, floating-point effects, and the work/depth trade-offs of each part.
Machine LearningOnsite - OpenAIHardMachine Learning EngineerImplement a 1-Nearest-Neighbor Classifier with a Manhattan-Distance Follow-Up
Implement a 1-nearest-neighbor classifier that labels each query point with the label of its closest training point under Euclidean distance, then extend it to Manhattan distance. Tests vectorized distance computation under memory limits, tie handling and how the choice of metric changes predictions.
Machine LearningTechnical Screen - OpenAIHardMachine Learning EngineerImprove Training With Noisy Annotators
This question evaluates a candidate's ability to handle label noise by estimating annotator reliability and applying data-cleaning, relabeling, or reweighting methods while measuring model performance and understanding classification metrics.
Machine LearningOnsite - OpenAIHardMachine Learning EngineerImprove classifier with noisy multi-annotator labels
This question evaluates a candidate's ability to analyze noisy multi-annotator labels, design label-aggregation and dataset-splitting strategies, and choose modeling and evaluation approaches for binary text classification.
Machine LearningTechnical Screen
Netflix
Nextdoor
OpenAI