Scaling Impact: Co-Creating a Shared Framework for Teaching AI and Machine Learning with Applications in Arctic Research

Artificial intelligence (AI) and machine learning (ML) play an increasingly crucial role in Earth science teaching and research by accelerating large data analyses, data interpretation, enhancing predictive models, and making complex environmental processes more accessible to students and researchers alike. The first NSF Cyber2A workshop, held in October 2024, was designed to provide Arctic researchers who have a limited background in computer science and computing with the tools and knowledge of AI and ML.

The 2025 workshop will bring together AI and ML practitioners, instructors, and other experts to collaboratively improve and expand upon this curriculum while designing an extensible framework for delivering training in different course formats. The expected outcome of the workshop is a modular set of instructional materials and supporting guidance that educators can use to teach useful AI/ML techniques to environmental and geosciences researchers, and that self-motivated learners can use themselves to develop useful skills directly. While the instructional examples will focus on Arctic research contexts, we invite applications from individuals across a range of disciplines, professional backgrounds, and career stages who are actively engaged in the application of AI and ML in their teaching.

Who Should Attend

This workshop is ideal for:

Faculty who are interested in offering AI-related courses centered on geoscience at their own institutions
Trainers who are interested in adopting and enhancing our course materials to provide AI training in their training programs
Applicants who have a solid foundation in Programming (Python, R, or similar) and AI & ML techniques

Workshop Objectives

Participants will:

Review foundational and advanced AI/ML applications relevant to geoscience research (e.g., feature detection using computer vision models, text analysis using large language models)
Develop pedagogical strategies for teaching AI/ML techniques to interdisciplinary research teams and students
Promote open science practices and ethical AI use
Foster a network of trained educators who can continue to build capacity within their institutions and regions
Collaborate on reviewing, designing, and creating reusable AI/ML training content

AI & ML Tools & Techniques Covered

We will delve into:

Principles and practices for preparing AI-ready training data
AI model building fundamentals
AI workflow validation, verification, and troubleshooting
AI model deployment for on-demand or batch prediction
Best practices in using tools and platforms such as Pytorch, MLOps, Tensorboard, and Tensorflow

Check out the curriculum from our 2024 workshop at https://cyber2a.github.io/cyber2a-course/.

Format

This is a highly collaborative and discussion-driven workshop. Expect a mix of presentations, small group activities, hands-on coding, and full room discussions.

Key Information

Dates: October 20-24, 2025
Location: In-person at The National Center for Ecological Analysis and Synthesis (NCEAS) in Santa Barbara, CA
Cost: Support is provided by the Cyber2A award from NSF and we have limited funds to support travel and accommodations for onsite courses.

Hear From Our Alumni

I found the Cyber2A workshop to be very helpful and particularly relevant to my research. Currently, I am exploring the capability of using Deep Learning U-net models to incorporate information about atmospheric NOx from satellites, reanalysis, and surface measurements to make predictions that could inform air quality forecasts. Our hope is to investigate whether machine learning could offer a less computationally expensive option for such forecasts at high resolutions.

- Mikhail Schee, Postdoctoral Fellow at the University of Toronto

Since attending, I’ve been developing a KMeans-based method to classify glacier surface conditions using Sentinel-2 imagery. This unsupervised approach has helped me label snow, ice, and land areas, which I then use to train supervised machine learning classifiers to estimate snow cover ratios and snowline altitudes across the Canadian Arctic.

- Wai-Yin Cheung

Project Showcase: Click an image to enlarge