Student-Built Dream Teachers: A Student-Led Agentic AI Framework for Personalized Learning

Twelve students aged 12–16 at Blue Blocks Montessori School, Hyderabad, designed and built their own personalized AI teachers from scratch — not as passive users but as active co-creators. Working through a five-section, 30-question preference-mapping instrument, each student identified their own emotional learning needs, desired teaching personality, communication style, and motivational preferences. They then translated these self-identified profiles into natural-language system prompts, integrated the Anthropic Claude API via Python, and deployed functional AI tutor prototypes on a web interface using Streamlit. The resulting "Dream Teachers" were individualized agents — not generic chatbots — each conditioned by its creator's own emotional and pedagogical blueprint. Across the diverse cohort, three overlapping themes emerged strongly: emotional safety, confidence-building support, and engaging, fun instruction. The paper presents the full preference-to-prompt pipeline, the layered code architecture, and the methodological significance of shifting students from passive AI consumers to active AI architects. It positions student-built agentic AI as a scalable and educationally meaningful model for future human-centred learning technology.