Graduate Teaching Assistants (GTAs) support a substantial share of undergraduate computing instruction, yet many enter their roles with limited preparation and opportunity for feedback. To create greater consistency and scalability, we implemented a redesigned GTA training at a large R1 university that integrates multiday workshops, asynchronous modules, and AI-simulated student communication practice. The model emphasizes professional communication, classroom management, and equitable teaching practices while generating diagnostic data to inform ongoing instructional coaching.

In two cohorts (n = 266), GTAs who completed workshops and simulations reported higher confidence and demonstrated stronger communication and adaptability performance than peers who completed online modules alone. These findings suggest that a structured, evidence-based orientation supported by AI student simulations shows promise for scalable implementation across large programs while maintaining pedagogical depth.