Amazon ML & AI interview questions.
Questions in the CrackML corpus associated with Amazon. Use company evidence as a prioritization signal, not a promise that an exact question will repeat.
How is BERT different from GPT?GenAI & LLM · medium · Evidence 88/100Explain Transformer architecture and each major component.GenAI & LLM · medium · Evidence 88/100How should dropout behave at test time?ML Fundamentals · medium · Evidence 82/100Explain the attention mechanism in a Transformer.GenAI & LLM · medium · Evidence 78/100Explain the bias-variance trade-off.ML Fundamentals · easy · Evidence 77/100Explain linear regression assumptions and failure modes.ML Fundamentals · medium · Evidence 77/100Walk through the toughest ML problem you solved and defend the approach.Project & Behavioral · hard · Evidence 77/100Implement Byte Pair Encoding.ML Coding & PyTorch · hard · Evidence 77/100When would you use RMSE instead of MAE?ML Fundamentals · medium · Evidence 77/100Explain PCA and when you would use it.ML Fundamentals · medium · Evidence 77/100Implement an in-place sorting algorithm and analyze every loop.Python & DSA · medium · Evidence 77/100Design an efficient sparse-matrix representation.ML Coding & PyTorch · medium · Evidence 77/100How do tree-based algorithms choose splits and generalize?ML Fundamentals · medium · Evidence 77/100How does multiclass logistic regression work?ML Fundamentals · medium · Evidence 77/100Design entity matching across different Alexa skills.ML System Design · hard · Evidence 77/100What does R-squared measure?ML Fundamentals · easy · Evidence 77/100How would you formulate and interpret R-squared for a regression system?ML Math · medium · Evidence 77/100When would you use a one-tailed vs two-tailed t-test?ML Fundamentals · medium · Evidence 77/100How does a multilayer perceptron learn nonlinear functions?ML Fundamentals · medium · Evidence 77/100Implement the transpose of a sparse matrix efficiently.ML Coding & PyTorch · medium · Evidence 77/100When would you choose gradient boosting over an SVM from a business perspective?ML Fundamentals · hard · Evidence 77/100Explain core NLP and deep-learning concepts used in an Alexa system.GenAI & LLM · medium · Evidence 77/100Explain binary logistic regression.ML Fundamentals · easy · Evidence 77/100Why and when would you ensemble models?ML Fundamentals · medium · Evidence 77/100How would you handle an imbalanced dataset?ML Fundamentals · medium · Evidence 76/100How do you calibrate a classifier?ML Fundamentals · medium · Evidence 75/100What is maximum-likelihood estimation?ML Fundamentals · medium · Evidence 75/100Walk through an ML project and defend your technical decisions.Project & Behavioral · medium · Evidence 75/100Write the mathematical form of Wasserstein distance and explain it.ML Math · hard · Evidence 75/100Explain regularization and why it works.ML Fundamentals · easy · Evidence 75/100How does an LSTM differ from a Transformer?GenAI & LLM · medium · Evidence 75/100How do linear and logistic regression differ?ML Fundamentals · easy · Evidence 75/100What is the Hoeffding inequality and when would you use it?ML Math · hard · Evidence 75/100Design a recommendation system for a product scenario.ML System Design · hard · Evidence 75/100Compare common loss functions and when to use them.ML Fundamentals · medium · Evidence 75/100What is the time complexity of matrix inversion?ML Math · medium · Evidence 75/100Explain attention in modern neural networks.GenAI & LLM · medium · Evidence 75/100How do you account for observed and unobserved confounding?ML Fundamentals · hard · Evidence 75/100Tell me about the most difficult technical project you handled.Project & Behavioral · medium · Evidence 75/100How do you reason about time and space complexity in ML-adjacent coding?Python & DSA · medium · Evidence 75/100Explain ResNet and why residual connections help.ML Fundamentals · medium · Evidence 75/100What are sparse-attention methods in Transformers?GenAI & LLM · hard · Evidence 75/100What does dropout do during neural-network training?ML Fundamentals · easy · Evidence 75/100Describe the Segment Anything Model (SAM) and how it worksML Fundamentals · medium · Evidence 75/100Explain Gaussian Mixture Models and their assumptions.ML Fundamentals · medium · Evidence 75/100When would you prefer gradient descent to the normal equation?ML Fundamentals · medium · Evidence 75/100How would you evaluate an LLM?GenAI & LLM · medium · Evidence 75/100How would you design an experiment when confounding is a concern?ML Fundamentals · hard · Evidence 75/100What is the closed-form solution for linear regression?ML Math · medium · Evidence 75/100How would you forecast daily grocery inventory for the next week?ML System Design · medium · Evidence 75/100Compare first-order and second-order optimizersML Fundamentals · hard · Evidence 75/100Design a RAG system end to end.GenAI & LLM · hard · Evidence 75/100How do DPO and RLHF differ?GenAI & LLM · hard · Evidence 75/100What is random about a Random Forest?ML Fundamentals · easy · Evidence 75/100Describe Segment Anything Model (SAM) at a high level.GenAI & LLM · hard · Evidence 75/100How is a split chosen in a decision tree?ML Fundamentals · easy · Evidence 75/100How would you model and evaluate time-series data?ML Fundamentals · medium · Evidence 75/100What role do activation functions play in neural networks?ML Fundamentals · easy · Evidence 75/100Design an A/B test for changing a buy button from green to yellow.ML Fundamentals · medium · Evidence 75/100Which losses are typical for linear vs logistic regression?ML Fundamentals · medium · Evidence 75/100Why use RAG instead of fine-tuning an LLM?GenAI & LLM · medium · Evidence 66/100Design a Sales-Based Ranking Data StructureML System Design · medium · Evidence 66/100Design Text-to-Image SearchML System Design · hard · Evidence 66/100Explain supervised fine-tuning for LLMs.GenAI & LLM · medium · Evidence 66/100Compare Random Forest and XGBoost.ML Fundamentals · medium · Evidence 63/100Compare bagging and boosting.ML Fundamentals · medium · Evidence 58/100How does KNN make predictions, and what breaks at large scale?ML Fundamentals · medium · Evidence 41/100How do normalization choices affect neural-network training stability and speed?ML Fundamentals · hard · Evidence 41/100Compare the inductive biases of LSTMs and CNNs for sequence modeling.ML Fundamentals · medium · Evidence 41/100How would you scope down an ambiguous ML system-design problem before modeling?ML System Design · medium · Evidence 41/100What data-quality problems would you check before modeling?ML Fundamentals · medium · Evidence 41/100When should domain adaptation be used?ML Fundamentals · hard · Evidence 41/100When does logistic regression struggle with feature interactions?ML Fundamentals · medium · Evidence 41/100How would you formulate an ML problem from a product goal?ML System Design · medium · Evidence 41/100When would you choose linear regression over a tree model for tabular prediction?ML Fundamentals · medium · Evidence 41/100How does Random Forest decorrelate its trees?ML Fundamentals · medium · Evidence 41/100How does knowledge distillation work and when would you use it?GenAI & LLM · medium · Evidence 41/100How would you choose between two plausible model families for the same objective?ML Fundamentals · medium · Evidence 41/100How do decision trees choose split thresholds?ML Fundamentals · medium · Evidence 41/100How would you choose evaluation metrics and explain them to a non-technical stakeholder?ML Fundamentals · medium · Evidence 41/100How would you choose a loss function for a speech or language-modeling task?ML Fundamentals · hard · Evidence 41/100Explain tokenization choices for an LLMGenAI & LLM · medium · Evidence 40/100How would you recognize and reduce overfitting?ML Fundamentals · medium · Evidence 40/100Explain reinforcement learning fundamentals for an Applied Scientist interviewML Fundamentals · medium · Evidence 40/100Compare SFT, RLHF, and preference optimizationGenAI & LLM · medium · Evidence 40/100How would you evaluate generative AI quality for a customer-facing product?GenAI & LLM · medium · Evidence 40/100Design text-to-image search with embeddingsGenAI & LLM · medium · Evidence 40/100Design an Image Classifier for 500 Tree SpeciesML System Design · hard · Evidence 34/100Backpropagation From First PrinciplesML Math · hard · Evidence 34/100Choosing a Loss FunctionML Fundamentals · medium · Evidence 34/100Evaluate a Computer Vision ModelML Fundamentals · medium · Evidence 34/100