Artificial Intelligence interview questions
500 questions with model answers, in four levels. Start with Basic, check yourself honestly, then move up.
Basic questions
- Math & Statistics for ML (13)
- Classical Machine Learning (13)
- Model Evaluation & Validation (13)
- Feature Engineering & Data Preparation (13)
- Deep Learning Fundamentals (13)
- CNNs & Computer Vision (12)
- RNNs & Sequence Models (13)
- Transformers & NLP (12)
- LLMs, Prompting, RAG & Agents (12)
- Generative Models (GANs, VAEs, Diffusion) (12)
- Reinforcement Learning (12)
- MLOps, Deployment & Responsible AI (12)
Intermediate questions
- Math & Statistics for ML (13)
- Classical Machine Learning (13)
- Model Evaluation & Validation (13)
- Feature Engineering & Data Preparation (13)
- Deep Learning Fundamentals (13)
- CNNs & Computer Vision (13)
- RNNs & Sequence Models (12)
- Transformers & NLP (12)
- LLMs, Prompting, RAG & Agents (12)
- Generative Models (GANs, VAEs, Diffusion) (12)
- Reinforcement Learning (12)
- MLOps, Deployment & Responsible AI (12)
Advanced questions
- Math & Statistics for ML (11)
- Classical Machine Learning (11)
- Model Evaluation & Validation (11)
- Feature Engineering & Data Preparation (11)
- Deep Learning Fundamentals (11)
- CNNs & Computer Vision (10)
- RNNs & Sequence Models (10)
- Transformers & NLP (10)
- LLMs, Prompting, RAG & Agents (10)
- Generative Models (GANs, VAEs, Diffusion) (10)
- Reinforcement Learning (10)
- MLOps, Deployment & Responsible AI (10)
Tricky questions
- Math & Statistics for ML (7)
- Classical Machine Learning (7)
- Model Evaluation & Validation (7)
- Feature Engineering & Data Preparation (6)
- Deep Learning Fundamentals (6)
- CNNs & Computer Vision (6)
- RNNs & Sequence Models (6)
- Transformers & NLP (6)
- LLMs, Prompting, RAG & Agents (6)
- Generative Models (GANs, VAEs, Diffusion) (6)
- Reinforcement Learning (6)
- MLOps, Deployment & Responsible AI (6)
Sample questions
- What are vectors and matrices used for in machine learning?
- What is principal component analysis (PCA)?
- What is data labeling, and why does label quality matter?
- What is the role of fully connected layers in a CNN?
- What is sentiment analysis?
- What is conditional generation?
- Why does gradient descent move in the direction of the negative gradient?
- How does ridge regression handle multicollinearity?
- How does SMOTE create synthetic samples?
- How does data augmentation improve a vision model, and what must you be careful about?
- How do you handle documents longer than a model's context window?
- How can generative models create synthetic training data, and what are the risks?
- An A/B test shows the new model lifts conversion from 4.0% to 4.3% with 20,000 users per arm. Is that significant?
- How do you evaluate a forecasting model fairly?
- How do you handle gradient explosion in a deep network?
- How would you build a document classification system for 50 categories with 20,000 labelled documents?
- How would you build an anomaly detector using a VAE?
- Does a correlation of 0 mean two variables are unrelated?
- Does label encoding categorical variables work for every model?
- Does fine-tuning always outperform prompting a strong model?