NASA Cosmic Origins Program

AI/ML Science and Technology Interest Group

Building AI Literacy for Astronomical Research

Mission Statement

The NASA Cosmic Origins Program AI/ML Science and Technology Interest Group (AI/ML STIG) addresses the critical need to upskill the astronomy community with AI literacy. We provide structured, domain-specific AI education through stackable, bite-sized modular training designed for astronomical research contexts.

About the STIG

The NASA Cosmic Origins Program AI/ML Science and Technology Interest Group (AI/ML STIG) addresses the critical need to upskill the astronomy community with AI literacy. We provide structured, domain-specific AI education through stackable, bite-sized modular training designed for astronomical research contexts.

Established under the Cosmic Origins Program Analysis Group (COPAG), the STIG brings together researchers and educators to build a comprehensive AI education framework tailored for the astronomy community.

Featured Lecture Materials

All lecture materials are open-source and designed to serve as templates for speakers and educators in the astronomy community.

Lecture 1: Overview

The Future of AI and the Mathematical and Physical Sciences

Jesse Thaler, MIT

An overview of how AI is transforming the mathematical and physical sciences, exploring opportunities and strategic priorities for the astronomy and astrophysics community. Based on the NSF Future of AI+MPS Workshop white paper.

Topics Covered:

  • AI's role in scientific discovery
  • Opportunities for astronomy & astrophysics
  • Building interdisciplinary AI+MPS communities
  • Strategic priorities and future directions
  • Education and workforce development
View Slides (PDF) Read White Paper on arXiv

Lecture 2: LLM API Basics

Large Language Models as Research Agents: Part 1

Yuan-Sen Ting, The Ohio State University

Learn the fundamentals of working with LLM APIs—making calls, managing conversations, and crafting effective prompts. Master key parameters, build multi-turn conversations, and implement prompting strategies for research tasks.

Topics Covered:

  • Understanding and using LLM APIs
  • Temperature, max_tokens, system prompts
  • Multi-turn conversations
  • Prompting strategies for research
  • Vision models for image analysis
View/Download on GitHub View Slides

Adapted from Coding Essentials for Astronomers.
Ting, Y.-S. (2025). Coding Essentials for Astronomers. Zenodo. DOI: 10.5281/zenodo.17850426

Lecture 3: Function Tools and RAG

Large Language Models as Research Agents: Part 2

Yuan-Sen Ting, The Ohio State University

Break through LLM limitations with function tools and Retrieval Augmented Generation. Build astronomical calculation tools, implement document-based Q&A, and create powerful research assistants.

Topics Covered:

  • Function tools for extending LLM capabilities
  • Astronomical calculation tools
  • Retrieval Augmented Generation (RAG)
  • Document chunking and embedding search
  • Vector databases for production systems
View/Download on GitHub View Slides

Adapted from Coding Essentials for Astronomers.
Ting, Y.-S. (2025). Coding Essentials for Astronomers. Zenodo. DOI: 10.5281/zenodo.17850426

Lecture 4: AI Agents

Large Language Models as Research Agents: Part 3

Francisco Villaescusa-Navarro, Flatiron Institute CCA

Explore autonomous AI agents that can reason, act, and collaborate. Learn about multi-agent systems, LangGraph workflows, and how AI agents are revolutionizing scientific research from literature search to experiment design.

Topics Covered:

  • AI Agents and ReAct (Reason + Act) framework
  • Multi-agent systems: collaboration & competition
  • Agentic workflows in scientific research
  • LangGraph for building agent systems
  • Applications: hypothesis generation, code generation, paper writing
View/Download on GitHub View Slides (PDF)

Lecture 5: Model Context Protocol

Large Language Models as Research Agents: Part 4

Yuan-Sen Ting, The Ohio State University

Introduction to the Model Context Protocol (MCP) for connecting LLMs to external data and tools. Learn how to build MCP servers to expose local resources and APIs to AI assistants.

Topics Covered:

  • Model Context Protocol (MCP) basics
  • Building MCP Servers
  • Connecting local data to LLMs
  • Resources and Tools
View/Download on GitHub View Slides

Adapted from Coding Essentials for Astronomers.
Ting, Y.-S. (2025). Coding Essentials for Astronomers. Zenodo. DOI: 10.5281/zenodo.17850426

Lecture 6: PyTorch & Automatic Differentiation

Deep Learning Frameworks: Part 1

Phill Cargile, Harvard-Smithsonian CfA

Master the fundamentals of PyTorch and automatic differentiation. Learn how to build, train, and optimize neural networks using PyTorch's powerful autodiff engine. From basic tensor operations to training your first neural network.

Topics Covered:

  • PyTorch tensors and operations
  • Automatic differentiation (autograd)
  • Building models with torch.nn.Module
  • Optimization with torch.optim
  • CPU vs GPU: device management
  • Training a minimal MLP from scratch
View/Download Notebook on GitHub View Slides (PDF)

Lecture 7: JAX

Deep Learning Frameworks: Part 2

Phill Cargile, Harvard-Smithsonian CfA

Dive into JAX, Google's high-performance numerical computing library. Learn how JAX combines NumPy-like syntax with automatic differentiation, vectorization, and JIT compilation to accelerate scientific computing and machine learning workflows.

Topics Covered:

  • JAX fundamentals and NumPy compatibility
  • Automatic differentiation with grad, value_and_grad
  • JIT compilation for performance
  • Vectorization with vmap and pmap
  • Functional programming paradigm
  • Building neural networks with JAX
View/Download Notebook on GitHub View Slides (PDF)

Program Schedule

Duration: 26-week series (November 1, 2025 - May 31, 2026)

Format: Weekly 1-hour sessions (40-45 mins + questions)

Time: Mondays at 4:00 PM ET

Delivery: Remote only

Meeting Link: Link can be found at NASA AI/ML STIG Official Page

Lecture Series Overview

Week Date Topic Speaker
1 Nov 3 Overview Jesse Thaler, MIT
Module 1: Large Language Models as Autonomous Agents
2 Nov 10 LLM API Basics Yuan-Sen Ting, OSU
3 Nov 17 RAG & Function Tools Yuan-Sen Ting, OSU
4 Nov 24 LLM as Agent Francisco Villaescusa-Navarro, Flatiron
5 Dec 8 Model Context Protocol (MCP) Yuan-Sen Ting, OSU
Module 2: Deep Learning Frameworks
6 Dec 15 PyTorch and Autodifferentiation Phill Cargile, Harvard-Smithsonian
7 Dec 22 JAX Phill Cargile, Harvard-Smithsonian
Module 3: Neural Network Theory
8 Jan 12 Inductive Biases John Wu, STScI
Module 4: Neural Network Architectures
9 Jan 19 Convolutional Neural Networks (CNNs) John Wu, STScI
10 Jan 26 Recurrent Neural Networks (RNNs) Daniel Muthukrishna, Harvard/MIT
11 Feb 2 Graph Neural Networks (GNNs) Tri Nguyen, Northwestern
12 Feb 9 Transformers TBD
Module 5: Physics-Inspired Networks
13 Feb 16 Equivariant Networks - Theory TBD
14 Feb 23 Equivariant Networks - Applications TBD
Module 6: Generative Models
16-20 Mar 2 - Mar 30 Normalizing Flows, Diffusion Models, Flow Matching, Simulation-Based Inference TBD
Module 7: Reinforcement Learning
21-23 Apr 6 - Apr 20 RL Fundamentals, Applications to Instrumentation & Telescope Scheduling TBD
Module 8: Data & Future
24 Apr 27 Open-Source Datasets and Best Practices TBD
25 May 4 Foundation Models for Astronomy TBD
Town Halls
26 May 11 Final Town Hall

Leadership Council

For inquiries, please contact: ting.74@osu.edu

How to Participate

Membership

The AI/ML STIG is open to the national and international community without regard to institutional affiliation, education, or career status. We welcome astronomers, astrophysicists, data scientists, and anyone interested in AI applications in astronomy.

Mailing List: Stay updated on upcoming lectures, events, and resources

Please send an email to AI-ML-STIG-join@lists.nasa.gov with the subject line "Join" to be added to the AI/ML STIG email list.

Recordings

All lecture recordings will be hosted on the NASA Cosmic Origins Program subpage, making them accessible to the broader community for asynchronous learning.

Resources & Links

GitHub Repository

Access all tutorial materials, code examples, and documentation

Visit GitHub

Community White Paper

Read our comprehensive white paper on AI literacy in astronomy

Read Paper

Cosmic Origins Program

Learn more about NASA's Cosmic Origins Program

Visit COR