About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Preparing Undergraduate and Graduate Students - and the Faculty who Prepare Them - for Materials Careers (The Judson Education Symposium)
|
| Presentation Title |
AskSparks: A Virtual Teaching Assistant Using Retrieval-Augmented Generation to Support Materials Science Students at Scale
|
| Author(s) |
Taylor D. Sparks |
| On-Site Speaker (Planned) |
Taylor D. Sparks |
| Abstract Scope |
This presentation introduces AskSparks, a retrieval-augmented generation (RAG) system designed to serve as a virtual teaching assistant in large materials science courses. Built on open-source LLM infrastructure, AskSparks transcribes and indexes over 700 YouTube videos—lectures and worked examples—from an Introduction to Materials Science and Engineering course. Students can ask natural language questions and receive targeted responses linked to timestamped video segments for deeper context and explanation.
By combining vector search over chunked transcripts with LLM-based answer synthesis, AskSparks delivers personalized, just-in-time support outside the classroom while reducing instructor and TA load. The system is scalable, course-adaptable, and designed for alignment with ABET outcomes related to lifelong learning and use of modern engineering tools. Initial deployment at the University of Utah demonstrates high student engagement and offers a blueprint for integrating AI into materials education. Implications for instructional design, faculty development, and workforce preparation will be discussed. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Sustainability, Other |