***New Section. Video coming soon*** 

Course Description:

Machine learning is transforming the way engineers tackle complex challenges — from predicting structural failures to designing new medical technologies. In this section, you’ll build a strong foundation in computational machine learning and discover how modern engineers use data-driven models to drive innovation.

You will learn how to represent complex data through vectorial embeddings — mathematical structures that allow machines to recognize patterns, measure similarity, and make decisions. A core theme of the course is understanding how the geometry of data — specifically, how points relate to one another in vector space — shapes a machine learning model’s ability to classify and predict accurately. You’ll practice creating and organizing training and test datasets, building classification models, and fine-tuning them based on embedding choices.

While we’ll work through real-world engineering applications, we’ll also dive into the underlying mathematics: exploring vector spaces, distance metrics, notions of similarity, and how they govern the performance of learning algorithms.

No prior machine learning experience is required. Some basic Python programming will be used, and support will be provided throughout the course.

Term Project:

In the second half of the course, you’ll work on a team-based project where you apply what you’ve learned to a real engineering problem. You’ll define a problem, collect or select appropriate data, create embeddings, train a model, and evaluate its performance. Example project ideas include:

• Predicting material failure based on sensor readings from mechanical tests

• Classifying environmental samples (e.g., water quality data) using chemical or biological measurements

• Detecting faults in power grids using time-series electrical data

Labs:  

Build embeddings with real-world data, construct and evaluate classification models, develop intuition about how the structure of data in vector space affects outcome, analyze how different measures of similarity (such as Euclidean distance and cosine similarity) impact machine learning decisions