ITA Interactional Competence Trainer

Each activity isolates one interactional skill — managing turns, checking understanding, redirecting off-topic students, handling confusion — and wraps it in a tightly scoped roleplay scenario. The AI student's behavior is scripted to create the exact moment where that skill is needed. After the session, a separate LLM reviews the full transcript and returns coaching feedback: one strength, one next step, and a skill status. Built with Next.js, LiveKit, and the OpenAI Realtime API. Developed in active collaboration with applied linguistics researchers at USF.