Machine Learning Foundations
For students, freshers and software engineers preparing for their first machine learning interview, with no prior ML experience assumed. You will be able to give clear, confident answers on core ideas such as supervised learning, data splitting, overfitting, bias and variance, feature scaling, evaluation metrics and regularisation, and back each one with a worked example.
Course Content
Prepares you to define machine learning, its main types and core terms such as model, training data and inference, in plain words with an example for each.
Prepares you to explain how learning with labels differs from learning without them, and to pick the right approach for a given business problem.
Prepares you to tell regression from classification by their outputs, loss functions and metrics, and to frame a real business problem as one or the other.
Prepares you to explain why data is split into train, validation and test sets, how leakage sneaks in through a bad split, and what to do when the dataset is small.
Prepares you to recognise overfitting and underfitting from train and validation scores or curves, explain why they happen, and name the right fix for each.
Prepares you to explain bias and variance with a clear picture, link them to underfitting and overfitting, and use the tradeoff to justify a choice of model.
Prepares you to explain why feature scaling matters, choose between normalisation and standardisation, and say which algorithms need it and which do not.
Prepares you to explain what features are, how to build and encode them from raw data, and which feature mistakes quietly break a model.
Prepares you to explain why models must be judged on unseen data, why accuracy can lie, and how to pick a metric that matches the business cost of each mistake.
Prepares you to read a confusion matrix, define the four outcomes without mixing them up, and explain when precision, recall or F1 is the right number to report.
Prepares you to explain how a probability becomes a yes/no decision, how moving the threshold trades precision for recall, and what the ROC curve and AUC really measure.
Prepares you to explain how L1 and L2 penalties stop a model from fitting noise, how they differ, how the strength setting controls complexity, and when regularisation makes things worse.
Prepares you to separate parameters from hyperparameters, explain grid and random search and why random usually wins, and describe how over-tuning quietly inflates your scores.
Prepares you to walk an interviewer through a structured debugging process — for poor scores, overfitting, and models that fail after deployment — and to name the mistakes that sink junior projects.