Five New CDS Electives: Spring 2027

What each new course covers, who it's for, and what you need before you register.

If you've ever wanted to build a language model instead of just prompting one, this is your semester. Spring's new electives take you inside how LLMs actually work, how platforms keep their users safe, and how AI is transforming weather and climate forecasting.

Two have no prerequisites, so they're open to students across the major. The other three build on core DS coursework and count toward applied and advanced methods requirements.

DS 592: Machine Learning for Inverse Problems

Instructor: Yves Atchadé

Prereqs: DS 122 or MA 581, and DS 320

Counts for: Applied and advanced

Some of the most important things in science can't be measured directly. A doctor can't see inside a patient without an MRI reconstructing an image from radio signals. Researchers can't place a sensor everywhere greenhouse gases are released. Recovering those hidden answers from limited, noisy data is called an inverse problem.

That's where DS 592 comes in. You'll learn to solve these problems with machine learning, and to quantify exactly how confident you can be in the result. Because inverse problems show up in medicine, climate science, astronomy, and engineering, the methods go with you into whatever field you care about.

Bonus: it's the only one of the five that counts toward both the applied and advanced requirements.

DS 593: Theory and Applications of Large Language Models

Instructor: Lauren Wheelock

Prereqs: None

Counts for: Applied use-inspired

Large language models like ChatGPT and Claude are a big part of why many students choose data science in the first place. DS 593 takes you inside the machinery. You'll start by building your own small transformer models from scratch, the architecture behind every modern LLM, so you understand exactly what's happening under the hood.

From there, the course gets hands-on. You'll apply pretrained models to real problems using fine-tuning, prompt engineering, retrieval-augmented generation (RAG), and AI agents. These are the same skills companies are hiring for right now, which makes this course a standout portfolio builder for anyone eyeing a career in AI.

With no prerequisites, it's one of the easiest ways to dive in.

DS 596: AI in Weather & Climate Simulation: From Days-to-Decades

Instructor: Elizabeth A. Barnes

Prereqs: None

Counts for: In-the-field

Weather forecasting is one of the most exciting frontiers in AI right now. Machine learning models can now produce forecasts that rival traditional physics-based simulations at a fraction of the computing cost. DS 596 shows you how those models are built, trained, and evaluated, from tomorrow's forecast out to simulations spanning decades.

You don't need to be a climate scientist to get a lot out of it. Forecasts drive decisions in energy, agriculture, insurance, supply chains, and public health, so the skills apply well beyond sustainability. And because it's a hands-on look at training and evaluating large models on real scientific data, it pairs naturally with DS 593 for anyone curious about how modern AI systems are developed.

And like DS 593, there are no prerequisites, so anyone can jump in.

DS 597: Methods for Trust & Safety

Instructor: Allison McDonald

Prereqs: DS 340

Counts for: In-the-field

Ever wondered how a scam gets caught before it reaches your feed? Behind every major platform is a team working to keep scams, misinformation, and harassment away from users. DS 597 shows you how that work actually gets done, through a mix of technical detection systems, legal frameworks, and human moderators.

That range is what makes the course stand out. You'll weigh the technical side alongside policy questions and the trade-offs platforms face in deciding where to spend limited resources. Whether you're headed toward engineering, policy, law, or business, trust and safety is a growing field that needs people who can work across all three.

DS 598: Graphs in Machine Learning

Instructor: Lisa Wobbes

Prereqs: Junior standing; DS 110, 121, and 320

Counts for: Advanced DS methods

In data science, a graph isn't a chart. It's a network: a set of things (nodes) and the connections between them (edges). Friends on a social app, molecules made of bonded atoms, roads linking cities, and transactions moving between bank accounts are all graphs, and some of the world's most valuable data looks like this.

DS 598 shows you how to unlock it. The course starts with graph mining, spectral methods, and embeddings, then moves to Graph Neural Networks, the models behind friend recommendations, fraud detection, and drug discovery. By the end, you'll be able to classify individual nodes (is this account a bot?) and whole graphs (is this molecule likely to be toxic?).

Overview

Not sure which one fits your degree plan? Your advisors can help!

View all posts