Unicon and Learnvia: A Practical Model for AI Math Tutoring
After months or years of hard work on a project, it’s always a proud moment when those efforts are recognized. In June, Learnvia, a free AI math tutoring program created by Carnegie Mellon University and developed in partnership with Unicon, was honored with an award for Teaching and Learning Innovation at the 2026 1EdTech Learning Impact Conference in San Francisco.
Learnvia brings together Carnegie Mellon’s learning science research and Unicon’s technical expertise in open-source platforms, interoperability, and scalable education technology. The platform uses adaptive instruction to help students succeed in challenging gateway math courses, a major barrier to degree completion in U.S. higher education.
The award in the AI-Augmented Learning Experiences category recognizes more than an innovative AI chatbot tutor. Across the education sector, foundations, institutions, and research organizations invest in promising ideas every year. Many demonstrate strong outcomes in pilots yet struggle to achieve broad adoption. Learnvia offers a sustainable model for adoption by combining decades of research, AI capabilities, interoperability, and operational sustainability to move from innovation to infrastructure.
Gateway Math as a Barrier to College Completion
In the United States, gateway courses derail the academic progress and degree completion for approximately 30% of learners across higher education. One subject in particular, introductory mathematics, is the greatest barrier to student success in higher education, according to the Mathematical Association of America.
For 15 years, the Gates Foundation and Carnegie Mellon have collaborated on courseware development, learning science research, and STEM education, identifying instructional approaches that improve persistence, comprehension, and mastery. As part of this commitment to accelerating learning and innovation, the foundation made a historic $55 million investment to develop Learnvia, focusing on Calculus 1 coursework.
But such projects often face obstacles like difficulty integrating with institutional systems, complex implementation requirements, and funding concerns after grants end. Clearly, success requires more than innovative technology. With a track record of shepherding educational technology concepts to the classroom, Unicon was brought in to develop Learnvia’s delivery platform and AI tutor incorporating Carnegie Mellon’s content and research.
Building an AI Experience Grounded in Learning Science
Many AI tutors are script-oriented, a highly labor-intensive process that results in what Christian Murphy, Unicon software architect and tech lead, calls “a choose-your-own-adventure game inside of a chat window.” At the same time, general-purpose AI chatbots are designed to answer prompts and follow a user’s request through to completion, even when that means doing the work a student needs to practice. “ChatGPT does not want you to struggle,” he says.
“We wanted it to feel a little bit more natural and interactive, but we didn’t want to lose the pedagogy. How do we teach this thing and make sure that the student really gets the challenge that they need to succeed?” — Christian Murphy, Software Architect and Tech Lead, Unicon
The team designed the AI tutor around the concept of productive struggle, the idea that students learn more effectively when they work through challenges with appropriate guidance rather than receiving immediate solutions. The team incorporated a computer algebra system that calculates the steps to reach the correct answer to a problem. Learnvia then builds a playbook for walking the student through the steps, and the AI tutor is built into the activity so students can call on it at any time. When they do, the tutor already knows what problem they are working on, what step they’re stuck on, and what they’ve already tried.
Lessons are scaffolded to guide students through problems, “never enough to be frustrated and quit, but enough that it sticks in their memory and they absorb the knowledge,” says Christian. This aligns with the concept of the zone of proximal development, which refers to the space between what a learner can accomplish on their own and what they can do with capable guidance. At this level, tasks are challenging enough to promote growth but not so difficult that learners become frustrated or disengaged.
The tutor’s context-awareness enables it to determine when to continue guiding the student versus providing the answer, with different levels of support for lessons and assessments. “We care about the progression of the student first and foremost,” says Christian. “They should be able to keep moving through the materials, so we have created the tutor personality to be friendly and encouraging.”
Designing an AI Math Tutor for Widespread Adoption
Designing an effective AI tutor was only part of the challenge. To ensure institutions would and could adopt the platform, Learnvia also had to align with existing education environments. This meant choosing a standard that aligns with common LMS infrastructure and reduces friction and complexity for students and instructors.
The Unicon team chose the Learning Tools Interoperability (LTI) standard for its widespread use in institutional LMS, especially its integration with Canvas. “Right there, we are meeting a majority of the market, where they just hit a button, and they are connected to our product,” says Christian. “You don’t need a new login, you don’t need a new grade book; it’s all there.”
These open standards and broad institutional accessibility have helped drive adoption: the Learnvia national network currently spans 38 institutions of higher education, from community colleges to four-year institutions. Calculus 2 is in the works, and plans are underway to expand to more quantitative courses.
A Model for Bringing AI Tutors to Reality
The significance of Learnvia’s award isn’t just the application of AI but the thoughtful and practical approach toward balancing technology, learning science, interoperability, implementation, and institutional realities.
As higher education institutions continue to explore the possibilities for AI in teaching and learning, Unicon continues to serve as a strategic partner, dedicated to moving these projects into sustainable operational reality. By focusing on the invisible infrastructure underpinning technological innovation, we help institutions transform pilot projects into core infrastructure positioned for long-term sustainability.
