Designing an LLM-Based IELTS Question Generator, Assessment, and Personalized Training System: Architecture and Research Agenda

While generative AI tools like LLMs have begun to assist in IELTS preparation, most fail to provide skill-specific, exam-aligned feedback across all four sections: Listening, Reading, Writing, and Speaking. This paper introduces NOAH, a novel LLM-based preparation system designed using a structured SDLC and LLMOps framework. Unlike generic AI tutors, NOAH features custom fine-tuned GPT-4o models trained on authentic IELTS materials, prompt-chained question generation, and automated visual and audio synthesis for realistic task simulation. The system supports modular skill pipelines, adaptive mock testing, and speech-based scoring using Azure APIs, delivering a full-stack solution optimized for personalization, scalability, and accessibility. Guided by human-centered AI principles, NOAH bridges pedagogical precision with generative AI flexibility, setting a foundation for equitable and high-fidelity language assessment preparation.