Looping the AI Lifecycle: Automated Training and LLM-Driven Evaluation via Tapis v3

This tutorial provides researchers with a comprehensive, hands-on introduction to advanced, reproducible Machine Learning (ML) workflows using the production-grade services of the Tapis v3 platform and ACESS System TACC Vista. Focusing on the complete machine learning engineering pipeline, participants will learn how to "loop" the AI lifecycle by orchestrating automated model training and executing sophisticated, LLM-driven evaluations directly on distributed computational resources. Through practical, step-by-step exercises, we will demonstrate how to seamlessly manage infrastructure using the Tapis Application Programming Interface (API) and the intuitive Tapis User Interface (UI). The session will deeply emphasize the integration of various Tapis core capabilities alongside specialized services, specifically launching Tapis jobs for automated training and FlexServ for LLM inferencing to build distributed reproducible scientific AI workflows. By the end of this tutorial, researchers and engineers will be fully empowered to efficiently develop, deploy, evaluate, and maintain automated AI lifecycles within their own research paradigms.