AI/ML interviews mix two things: classical machine learning theory — the kind you can be quizzed on a whiteboard — and the modern stack built around Python, PyTorch or TensorFlow, and increasingly transformers and LLMs. Below are the questions asked most often, each with a model answer. Junior: core theory, definitions, and simple practical cases.
Supervised learning trains on labeled input-output pairs, like predicting house prices from features. Unsupervised learning finds structure in unlabeled data, like clustering customers into segments. Reinforcement learning trains an agent to take actions that maximize a reward signal over time, like a game-playing bot — there are no labeled examples, only feedback from the environment.
A model that overfits has learned the training data too well, including its noise, so it performs great on training data but poorly on new data. A classic sign is a big gap between training accuracy and validation accuracy. The fix usually starts with getting more data or simplifying the model before reaching for fancier regularization.
A loss function measures how wrong a model's prediction is compared to the true value — mean squared error for regression, cross-entropy for classification. Training is literally the process of adjusting the model's parameters to make this number smaller. The choice of loss function shapes what the model actually optimizes for, which is why picking the wrong one silently produces the wrong model.
Precision is: of everything the model flagged as positive, how much was actually positive. Recall is: of everything that was actually positive, how much did the model catch. A spam filter tuned for high precision misses fewer real emails but lets more spam through; tuned for high recall it catches more spam but flags more real emails too — there is always a trade-off.
NumPy provides fast array operations in Python, working on data as vectorized arrays instead of slow Python loops — nearly every ML library is built on top of it. Pandas builds on NumPy to give you DataFrames — labeled, tabular data you can filter, group, join and clean, which is where most real-world data work actually happens before a model ever sees the data.
In a real interview you speak under pressure. Cam asks these same questions, scores every answer, and shows exactly what to fix.
Practice a Junior AI/ML Engineer interview →