Machine Learning Fundamentals
How does machine learning differ fundamentally from traditional software engineering?
What is the conceptual distinction between artificial intelligence, machine learning, and deep learning?
How would you explain what machine learning is to a non-technical executive?
What are features, labels and instances in a machine learning dataset?
Can you walk me through the end-to-end machine learning workflow from problem framing to deployment?
What is the difference between a machine learning algorithm and a machine learning model?
When should you not use machine learning to solve a problem?
What is the conceptual distinction between machine learning and classical statistics?
Why does the phrase "garbage in, garbage out" matter more in machine learning than in traditional software?
What is the difference between supervised and unsupervised learning?
What is semi-supervised learning?
What distinguishes reinforcement learning from supervised learning?
How does self-supervised learning differ from standard unsupervised learning?
How does a generative task such as text or image generation differ from a predictive task like classification or regression?
What is the difference between transductive learning and inductive learning?
When would you frame a problem as clustering rather than classification?
When would you frame an application as a ranking task rather than a classification task?
When would you frame a problem as reinforcement learning instead of supervised learning?
How do you define what success looks like for an ML project before you build a model?
How do you frame an ambiguous business goal as a well-defined machine learning task?
How do you choose a target variable when the true business outcome cannot be observed directly and you must rely on a proxy label?
How would you decide between building one global model and several segment-specific models?
Generalization & Inductive Bias
What is the difference between parametric and non-parametric machine learning models?
What is the i.i.d. assumption in machine learning?
What is generalization error?
What is the difference between underfitting and overfitting?
Why does high training performance fail to guarantee real-world success?
How does Occam's razor apply to machine learning model selection?
What is a hypothesis space in machine learning?
What is the inductive bias of a linear regression model?
What is the inductive bias of a decision tree?
What are the practical engineering implications of the No Free Lunch theorem?
Why is inductive bias necessary for a model to generalise to unseen data?
Why does high correlation between a feature and the target not necessarily imply a causal relationship?
Why is class imbalance a fundamental problem for standard loss functions?
How does label noise in the training set affect a model's learned decision boundary?
Why does iterating on data quality often produce better returns than tuning model architectures?
How would you obtain training labels when labelled data is scarce or expensive to collect?
How would you approach an ML problem where you have only a few hundred labelled examples?
What is the difference between training and inference in machine learning, and why do their costs differ?
Why should you generally start with a simple model before trying complex architectures?
What is the tradeoff between model interpretability and predictive performance?
When is a deep learning approach justified over classical machine learning algorithms?
How do latency and compute constraints influence model selection?
How do you choose between a linear model, a tree-based ensemble, and a deep neural network for a given problem?
When would you choose an instance-based model like k-NN over a model-based algorithm?
What are the main families of machine learning models, and what is each one characteristically good at?
Baselines & Error Analysis
What is a heuristic baseline?
What constitutes a good baseline for a classification task?
What constitutes a good baseline for a time-series or regression task?
How does error analysis guide model iteration?
How would you walk an interviewer through an ML project you built, end to end?
How would you decide whether a model is good enough to ship to production?
Production & Deployment Risks
What is training-serving skew?
What are the primary sources of bias and unfairness in machine learning systems?
Why does a deployed model need ongoing monitoring and periodic retraining?
What is a feedback loop in production ML systems?
What happens to a model when the i.i.d. assumption is violated in production?
What is the difference between concept drift and covariate shift?
What are the typical risks of deploying an ML solution compared to a deterministic heuristic?