All interview questions
1501 questions across 7 topics. Each one has its own page with a full answer and a multiple-choice check.
A sample of each topic is listed below. Follow “All questions” to see the complete list for that topic.
- What is the bias-variance tradeoff?
- How do you detect and fix overfitting?
- Precision vs recall — when do you optimize each?
- When should you use PR-AUC instead of ROC-AUC?
- How do you handle class imbalance in a dataset?
- L1 vs L2 regularization — what's the difference?
- What is k-fold cross-validation and when do you use stratified or grouped folds?
- Why is logistic regression called a linear model if it uses a sigmoid?
- Random forest vs gradient boosting — which do you pick and why?
- Which models need feature scaling and which don't?
- What is target leakage and how do you prevent it?
- Generative vs discriminative classifier — what's the difference?
- How does k-means work and what are its main limitations?
- How do you choose the number of clusters k?
- When would you pick DBSCAN over k-means?
- PCA vs t-SNE vs UMAP — when do you use each?
- How do you decide how many PCA components to keep?
- What does the silhouette score measure?
- What is hierarchical clustering and when is it useful?
- What are the main approaches to anomaly detection?
- Gaussian Mixture Model vs k-means — what's the difference?
- What is self-supervised learning and why does it matter?
- How does contrastive learning work?
- What is an autoencoder and what are the common variants?
- ReLU vs sigmoid vs GELU — when do you use each?
- What are vanishing and exploding gradients, and how do you fix them?
- What does Batch Normalization do?
- When do you use LayerNorm instead of BatchNorm?
- How does dropout work and when is it applied?
- SGD vs Adam vs AdamW — how do you choose?
- Why use a learning-rate schedule (warmup + cosine decay)?
- What inductive biases do CNNs have?
- What is the receptive field in a CNN?
- Why do residual (skip) connections help training deep networks?
- How does self-attention work in a transformer?
- Why does a transformer need positional encoding?
- Define the RL setting: agent, environment, state, action, reward, policy.
- What is the exploration-exploitation tradeoff?
- On-policy vs off-policy learning — what's the difference?
- What update does Q-learning perform?
- What is a policy gradient and why do we use it?
- What is an actor-critic algorithm?
- What does the Bellman equation say?
- What role does the discount factor gamma play?
- What are the key tricks that make DQN work?
- Why is PPO so widely used?
- What is reward shaping and what are its pitfalls?
- What is the credit assignment problem?
- What is a large language model, in one paragraph?
- Why do LLMs hallucinate and how do you reduce hallucinations?
- How does Retrieval-Augmented Generation (RAG) work?
- What are text embeddings and how are they used?
- How should you chunk documents for RAG?
- What do temperature and top-p do at generation time?
- What are the pillars of good prompt engineering?
- Fine-tuning vs RAG — which do you use when?
- What is LoRA and why is it popular for fine-tuning LLMs?
- What is the context window and what tricks extend it?
- What is a Mixture-of-Experts (MoE) LLM?
- How do you evaluate an LLM system?
- What is data drift and how do you detect it?
- How is concept drift different from data drift?
- What does 'data imbalance' mean and why is it a problem?
- What are common sources of data leakage in an ML pipeline?
- What is training-serving skew?
- What problem does a feature store solve?
- What should you monitor in a production ML model?
- What is a shadow deployment and how does it differ from A/B testing?
- What is a canary release for ML models?
- When and how do you retrain a production model?
- What does CI/CD look like for ML?
- How do you validate data quality in an ML pipeline?
- State the Central Limit Theorem in one sentence.
- What is a p-value, precisely?
- Type I vs Type II error — what's the difference?
- What does a 95% confidence interval mean?
- Bayesian vs frequentist — what's the core difference?
- Why doesn't correlation imply causation?
- What is the bootstrap and when do you use it?
- What is Maximum Likelihood Estimation?
- State the Law of Large Numbers.
- Independence vs uncorrelatedness — what's the difference?
- State Bayes' theorem and one intuitive use.
- Why do we use standard deviation instead of variance in interpretation?