Syllabus
Course description
Machine Learning in Practice is a project-based course that gives students training and experience solving real-world problems with machine learning, exploring the interface and gaps between research and practice. We focus on the nuance of using ML in the real world, where common assumptions (like i.i.d. and stationarity) break down, and where there is a growing need for approaches that go beyond optimizing for accuracy: fairness, explainability, robustness, uncertainty quantification, and real-world impact. Through lectures, discussions, presentations, readings, and a semester-long team project, students build ML systems for real problems with real data.
Learning objectives
- Scope & formulate. Translate a messy real-world problem into a well-posed ML task (predictive or generative), decide whether ML is even the right tool, and define success in terms of actions, decisions, and outcomes rather than offline metrics.
- Build end-to-end systems. Design and implement modular, config-driven ML pipelines and foundation-model systems on real, messy data.
- Validate Choose validation strategies that match how the system must generalize, and detect the ways both predictive and generative systems silently fail.
- Explore beyond accuracy. Learn about, apply, and critically evaluate methods for fairness, interpretability, uncertainty, robustness, and causality.
- Evaluate impact. Move from offline metrics to exploring field evaluation of whether a system actually improves real-world outcomes.
- Communicate. Present technical work clearly to technical and non-technical audiences, in writing and in person.
How the course is organized
| Module | Weeks | Focus |
|---|---|---|
| 1 · Building ML Systems | 1–6 | Scoping, formulation, data, validation, metrics, features, pipelines |
| 2 · Beyond the Basic Model | 9–14 | Building with foundation models & agents; evaluating generative systems; ethics, causality, field evaluation, robustness, uncertainty, interpretability, fairness |
The applied ML project runs throughout the semester. See the schedule for the week-by-week plan with readings, and the project page for the milestone timeline.