We help students learn math by teaching it.
MELT-Math flips the classroom dynamic. Instead of being tutored by AI, students teach an AI learner — explaining, correcting, and reasoning out loud. The result is deeper understanding, durable confidence, and a love of mathematics.
Our vision
A world where every learner builds deep understanding and applies it beyond the classroom.
When students teach, they surface misconceptions, explain their reasoning, and refine their mental models. That process builds deep conceptual understanding, not just the ability to get the right answer.
And because understanding is built through explanation and reflection, students can transfer what they learn to authentic tasks, problem solving, and real-world contexts.
Decades of learning science, distilled.
MELT-Math didn't start as a product. It started as a question in the learning sciences: why do people understand things more deeply once they have to teach them?
The foundation
Research on learning-by-teaching shows that explaining ideas to others reorganizes a learner's own understanding.
The protégé effect
Studies confirm students invest more and learn more when responsible for a pupil's success — even a simulated one.
MELT-Math
We pair that science with modern AI: a teachable agent students guide, correct, and learn alongside.
Educators, scientists, and builders
Dr. Wanli Xing
CEO & Founder
John S. and James L. Knight Foundation Chair in Artificial Intelligence at the University of Miami, Director of the Advanced Computing for Education Lab, and a nationally recognized leader in AI-powered learning. His research, supported by the NSF, IES, ONR, NIST, and Gates Foundation, spans more than 6,100 scholarly citations and over $30M in competitive federal funding.
Dr. Zeyu Yan
AI and Development Lead
Ph.D. in Mechanical Engineering from The University of Texas at Austin, with more than 20 research publications spanning machine learning, large language models, and intelligent systems. He leads the engineering and AI infrastructure behind MELT-Math, turning cutting-edge research into scalable educational technology.
Dr. Jihyun Rho
Learning Sciences Lead
Ph.D. in Learning Sciences and researcher specializing in AI in education, learning analytics, and visual learning. She leads the pedagogical design of MELT-Math, ensuring every AI interaction is grounded in evidence-based learning science and supports deep conceptual understanding.
Dr. Lori Dassa
School Partnership Lead
Associate Professor of Practice and educator with extensive experience in curriculum, teacher professional learning, and K–12 school partnerships. She leads MELT-Math's implementation strategy, helping research-backed AI translate into meaningful classroom practice at scale.
Jiajun Chen
UI/UX Lead
Product designer and design manager with nearly a decade of experience crafting mobile, web, and desktop applications across human factors and user-centered design. He leads the product design of MELT-Math — from research and strategy to high-fidelity UI — ensuring the learning-by-teaching experience is intuitive, engaging, and accessible to every student.
Research-backed. Classroom-validated.
1.5
Effect size by research
Evaluated through three classroom design iterations with middle school learners.
533+
Student-AI interactions analyzed
Research identified engagement patterns associated with stronger mathematics learning.
3000+
Students researched
Grounded in four decades of learning-science research on how students learn by teaching.
Supported by leading research & innovation organizations
Three priorities driving what's next
Our roadmap is built on three strategic pillars — making MELT-Math more expressive, more trustworthy, and within reach of every classroom.
Multimodal
Multimodal learning
Teach with diagrams, visuals, speech, and mixed representations — so students can reason across subjects, not just text.
Responsible
Responsible AI
Privacy, safety, and transparency by design. We mitigate bias and keep humans at the center of every interaction.
Scalable
Affordable at scale
Efficient and low-cost by design, so any classroom can adopt learning-by-teaching — not just the well-resourced ones.
Join us in transforming mathematics learning
See how learning-by-teaching works in your classroom.