AWS Public Sector Blog

Personalized learning support at scale: How UT Austin built a generative AI tutor platform on AWS

AWS Branded Background with text "Personalized learning support at scale: How UT Austin built a generative AI tutor platform on AWS"

As generative artificial intelligence (AI) becomes a prominent tool in students’ educational journeys, universities are racing to adapt while safeguarding academic integrity. A recent study from Harvard’s Graduate School of Education found that more than half of US teens and young adults (53 percent) have used generative AI, primarily for seeking information and brainstorming. As these tools become embedded in how students learn and study, institutions must determine how to manage these tools responsibly to support accuracy and alignment with academic values.

Building on a commitment to using innovative technologies to enhance the learning experience, the University of Texas at Austin (UT Austin) collaborated with Amazon Web Services (AWS) to develop UT Sage, a faculty-guided, generative AI tutor platform. UT Sage provides conversational course-related support on-demand while aligning with responsible AI frameworks and preserving the essential connection between faculty and students.

In this post, learn how UT Austin worked with the AWS Generative AI Innovation Center to build UT Sage and what this experience can teach other institutions navigating the promise and responsibility of AI in higher education.

Deepening student learning with generative AI course tutoring

UT Austin is one of the largest public universities in the US, serving approximately 52,000 students across 19 colleges and schools. With a strong focus on academic innovation and excellence, university leaders recognized both the opportunities and challenges of generative AI as these tools gained popularity among students.

“We wanted to broaden access to AI tools on campus in a way that aligned with established learning sciences and our responsible AI adoption framework,” said Kasey Ford, AI designer at UT Austin. “Our goal was to create a platform that promoted evidence-based pedagogical practices that faculty could trust and that students could use to support their academic goals.”

These conversations resulted in UT Sage, a generative AI-powered tutor platform that allows faculty to create customized virtual tutors aligned with their course material. UT Sage’s AI tutors engage students through Socratic dialogue, promoting durable skills like critical thinking rather than simply retrieving answers. Faculty can create tutors for various purposes, such as reinforcing prerequisite or foundational knowledge, deepening understanding of course topics, and more.

“For the students, the tutor works a lot like the kind of chatbot everyone is familiar with, but the advantage is that it’s been trained by their instructor to align with what’s being taught in their courses,” said Ford, who also serves as product owner for UT Sage.

For faculty, UT Sage offers an efficient way to extend learning beyond the classroom. “The UT Sage project empowers our faculty to create personalized learning tools, designed to motivate student engagement,” said Julie Schell, assistant vice provost and director of the Office of Academic Technology. “With the potential to deploy across hundreds of courses, we are aiming to enhance learning outcomes and reduce the time and effort required to design student-centered, high-quality course materials.”

Collaborating with the AWS Generative AI Innovation Center 

Scalability was particularly important as the UT Austin team started planning this solution. The university wanted a solution flexible enough to serve a wide variety of disciplines, from chemistry to philosophy. Manually creating personalized tools would have been too time-consuming at scale, and in-house expertise to design and build the kind of agentic AI system it envisioned was limited.

To bring Sage to life, UT Austin turned to the AWS Generative AI Innovation Center, an acceleration team at AWS that connects organizations with AWS AI and machine learning (ML) experts to help ideate, build, and deploy generative AI solutions.

“One of the things that made working with the AWS Generative AI Innovation Center so impactful was that it wasn’t just high-level advising—it was hands-on-keyboard support,” said Ladd Hanson, enterprise cloud architect at UT Austin. “They were writing code, helping with solution design, and meeting with us regularly to stay aligned so we could move quickly.”

AWS team members, in turn, praised UT Austin’s commitment to real-world user testing. “The user testing informed a lot of the features we added and made this a very robust application,” said Robby Milletich, senior applied scientist at AWS. “The UT Austin staff set the bar very high in terms of user testing.”

Building a scalable, faculty-guided generative AI tutor platform

 UT Sage is designed to create a seamless generative AI tutoring experience for both instructors and students. Faculty follow established learning science principles when they use UT Sage to create topic-specific AI tutors aligned with their courses, and students interact with these tutors in a natural, conversational interface that fosters deeper engagement with class material.

The AWS architecture supporting Sage includes:

UT Sage uses a lightweight agentic AI approach that dynamically accesses a knowledge base of course material and relevant tools to answer student questions. For example, if a student asks, “When is the test?” or “When is an assignment due?” UT Sage AI accesses the syllabus for answers. If a student inputs into UT Sage, “I can’t understand the assignment” or “I want to dive deeper into this topic” the AI pulls from faculty-approved course material. When appropriate, UT Sage prompts students in Socratic-like dialogue to help them think critically and reflect instead of simply providing direct answers.

Security and responsible AI use were key development priorities in the Sage project. Each topic tutor is grounded in curated, instructor-provided course materials, with system prompts carefully designed to maintain domain focus and prevent off-topic or inappropriate responses.

Lessons learned from building a university-wide generative AI tutor

 After about a year of ideation, development, and testing, UT Sage is currently in open beta. So far, early feedback from students and instructors has been overwhelmingly positive. “Anecdotally, we’re hearing from instructors and students alike that they are excited to use the tool,” said Ford. “Even those who were AI reluctant, or who understood the limitations of AI to help with studying, for example—even those students are excited to use tutors that have been trained directly by their instructors.”

As advice for other universities or academic institutions looking to build generative AI solutions, UT Austin stresses the importance of listening to the needs of students and faculty first. “It’s critical to understand the pedagogical purpose or value of adopting any technology, versus latching onto a particular platform because it sounds cool or has lots of buzz and then trying to find a suitable use case for it,” suggested Ford.

UT Austin also credits the project’s success to several other key practices:

  • Extensive UI/UX design upfront: Before coding began, the team invested months in user experience and interface design to make sure the application would be intuitive and supportive for students and faculty.
  • Rigorous user testing: UT Austin engaged real student and instructor testers early and often, refining Sage’s dialogue styles and functionality based on direct feedback.
  • Responsible AI frameworks: UT Austin maintained transparency with stakeholders and aligned UT Sage’s development with campus-wide responsible AI and governance initiatives, so the application stays secure, focused, and accurate.

Looking ahead: Enhancing learning and human connection with AI tools

UT Sage is scheduled for a broader rollout across the university in fall 2025, with planned integrations into UT Austin’s learning management system (LMS). Upcoming additional features aim to enhance instructor insights, like a dashboard that shows aggregate data on the questions students ask to help inform in-class teaching.

“Our ultimate hope for any technology adoption is that it enhances the learning experience between faculty and students, that it doesn’t replace those relationships,” added Ford. “We hope UT Sage will make face-to-face teaching and learning even more effective.”

“We believe UT Sage may be one of the first scalable, research-based virtual instructional designers in higher education,” noted Schell. “This collaboration with AWS marks a significant step in our journey to accelerate education in the AI era.”

To learn how AWS helps institutions build and launch transformative cloud solutions to support students and campus operations, visit AWS for Higher Education or contact us today to get started.

Read related stories on the AWS Public Sector Blog

TAGS: , , , ,
Eric Weigel

Eric Weigel

Eric is the senior technology manager for The Emerging Technologies and Architecture Team at The University of Texas at Austin. His team is responsible for supporting the university’s cloud infrastructure and services. The team is developing the UT Sage application which is a generative AI tutor bot platform for faculty and students.

Amritha Shetty

Amritha Shetty

Amritha is a senior solutions architect at AWS. She works with public sector customers, guiding them through their AWS cloud journey. She brings over fifteen years of software design and development experience.