This course gives you a deep, end-to-end practical understanding of machine learning. You will frame ML problems, prepare data, train and evaluate classical and deep models, deploy them, and monitor them in production. Across the curriculum, you will use scikit-learn for classical ML, XGBoost/LightGBM for tabular gold-standards, and PyTorch for deep learning. You will master the complete project lifecycle: define the metric, baseline, iterate, validate, ship, and monitor.
The modules cover a wide array of technical subjects including feature engineering with numerical, categorical, text, and time-series data. Classical model training explores linear and logistic regression, tree-based models, XGBoost, LightGBM, CatBoost, KNNs, and SVMs. Unsupervised learning modules dive into clustering algorithms (K-means, DBSCAN) and dimensionality reduction techniques (PCA, t-SNE, UMAP), alongside recommender systems.
Deep learning essentials provide hands-on experience building MLPs, CNNs using transfer learning (ResNet/EfficientNet), and NLP Transformers fine-tuned via Hugging Face. Advanced topics include model evaluation, calibration, explainability via SHAP and LIME, bias mitigation, and hyperparameter tuning with Optuna. The program culminates in MLOps training, teaching students to package models using FastAPI, ONNX, and Triton, manage CI/CD pipelines, and monitor for concept drift using Evidently and WhyLabs.
