Episode Overview

Artificial Intelligence is rapidly transforming nearly every profession, yet many educational institutions are still searching for effective ways to integrate AI into teaching and learning.

In this episode of Learning with AI, host Gladys Mercier sits down with David Miller, founder of ScholarStack and instructor of AI Engineering at Carnegie Mellon University, to discuss how AI can be intentionally designed to enhance learning experiences for students, support faculty, and provide meaningful insights for institutional leaders.

Their conversation explores the intersection of learning science, educational technology, and the future of higher education.

Why ScholarStack Was Created

David's journey to building ScholarStack began with a simple observation. As AI tools emerged across industries, specialized solutions quickly appeared for lawyers, software engineers, and healthcare professionals. Yet education lacked a platform specifically designed around how people learn.

Rather than adapting general-purpose AI tools to classrooms, ScholarStack was built as an AI-native learning platform designed specifically for educational environments. Its goal is to help educators create structured, personalized learning experiences while giving institutions greater visibility into how learning is happening across courses and programs.

"AI is incredibly powerful, but it wasn't built specifically for educational needs. Our goal is to harness that power and deliver it in a way that is intentionally designed for teaching and learning."

Learning Science at the Center

One of the key themes of the conversation was the role of learning science. According to David, decades of educational research have already shown what effective learning looks like. The challenge is that individualized instruction—the gold standard of learning—is often difficult to scale.

ScholarStack attempts to bridge that gap by allowing instructors to create guided AI experiences that reflect their teaching style, course materials, and learning objectives. These experiences can take the form of structured conversations, coaching sessions, simulations, or guided practice activities that help students engage more deeply with course content.

What Students Experienced

During the Spring semester, David introduced ScholarStack in his AI Engineering course at Carnegie Mellon University. One piece of student feedback stood out.

After completing a guided interview exercise designed to reinforce course concepts, a student expressed frustration—not because the activity was difficult, but because it ended. She wanted to continue.

For David, that response represented something every educator hopes to see: genuine engagement. Rather than viewing learning activities as tasks to complete, students were actively seeking more opportunities to explore and apply what they had learned.

Giving Faculty Greater Visibility

The conversation also explored how AI can help instructors better understand student learning. Through ScholarStack's analytics and observability tools, faculty can see:

  • How students are interacting with learning experiences
  • Areas where students may be struggling
  • Patterns of engagement across an entire class
  • Emerging themes, questions, and misconceptions

Instead of reviewing hundreds of individual interactions manually, instructors can receive summarized insights that help them identify where intervention or clarification may be needed. This allows educators to spend less time searching for problems and more time supporting learning.

Supporting Institutional Goals

Beyond individual courses, ScholarStack is designed to serve administrators and academic leaders. Institutions often invest significant effort into demonstrating that learning objectives are being achieved across departments, programs, and colleges. The platform helps surface evidence that can support:

  • Learning outcome assessment
  • Program evaluation
  • Institutional improvement initiatives
  • Accreditation preparation

By organizing and analyzing educational activity at scale, administrators gain access to insights that are often difficult—or nearly impossible—to collect manually.

Working with Existing AI Ecosystems

Many institutions have already established relationships with providers such as ChatGPT, Claude, or Gemini. Rather than requiring schools to replace those investments, ScholarStack can integrate with existing AI infrastructures.

This flexibility allows institutions to continue using their preferred AI providers while creating educational experiences specifically designed for teaching, learning, and assessment. For technology leaders, the conversation often centers on security, privacy, governance, and return on investment—all areas that have become critical considerations in AI adoption.

Preparing Students for an AI-Powered Future

The episode concludes with a broader question: how should colleges and universities prepare students for an AI-saturated workforce?

David argues that the answer is not simply giving students access to AI tools. Instead, institutions must focus on creating meaningful learning experiences that help students think critically, practice problem solving, and engage deeply with ideas while leveraging the capabilities of AI.

The opportunity is not to replace learning with technology, but to use technology to make learning more effective, more personalized, and more engaging.

Key Takeaways

  • AI should be intentionally designed around learning science.
  • Personalized learning experiences can become more scalable through AI.
  • Faculty benefit from deeper visibility into student learning.
  • Administrators gain access to institution-wide educational insights.
  • AI adoption in education requires thoughtful integration, not just tool deployment.
  • The future of higher education will depend on how effectively institutions combine human teaching expertise with AI capabilities.


Episode Transcription

Gladys Mercier: hello everyone. Today we are lucky to have David Miller who is the founder of Scholar Stack. David, welcome.

David Miller: Thank you. Wonderful to be here. I'm excited for the conversation. Gladys

Gladys Mercier: Absolutely. I'd like to start by maybe having you tell us a little about what is Scholar Stack and uh your background, how you got interested in uh creating this fantastic

David Miller: Great.

Gladys Mercier: tool.

David Miller: Yeah, happy to start there. So, Scholar Stack is an AI native learning platform that uh helps teachers design learning experiences, AI powered learning experiences for students. Um, basically in the past few years as I watched AI proliferate, I sat back and wondered where was the solution for the education vertical and uh, you know, lawyers had Harvey uh, Meta Medicine had um, Socratic um, you know, software engineering had a few of them specifically tailored for software engineering.

David Miller: and I said, "Where is the one that should be specifically tailored for um for higher ed?" And I and I didn't see it. Um and so I set out to tinker and uh I I I kind of relate it to to like a home renovation when you think, "Oh, I'm just gonna just gonna put in some new overhead lights. No big deal." And then you ultimately decide you want to tear down the wall and raise the roof and put in tile floor. And so over the past year, Scholars Tech has really grown in terms of platform to include a very deep and rich functionality that serves the professor or the teacher, the student as well as uh university or school administration and the IT organization at the school. So, it's a comprehensive uh I uh AI infrastructure platform for schools that is really centered on the learning experience. Uh and and the goal uh at Scholar Tech is to really harness AI to deliver a um an intentional um deliberate, structured and enjoyable learning experiences for students that can be uh tailored and customized and designed uh by the professor or the teacher.

Gladys Mercier: Excellent. Excellent. So you were an instructor as well as a founder and this is where

David Miller: Yes.

Gladys Mercier: the idea came from.

David Miller: Yeah.

Gladys Mercier: Yeah.

David Miller: I mean, I've been teaching for about 10 years. Um, and but for most of my career, I've been a product management professional. So, building and shipping products and understanding the user experience. Um, and and I'm also teaching AI engineering at Carnegie Mellon. And so those three areas really blended together perfectly to give me the kind of background and the bird's eye view as well as the practitioner understanding of how things get built and how to um you know build user- centered design products and um and so yeah the my background has informed the platform greatly and um and I I debuted the platform form in my class this spring and any any entrepreneur or product person knows you think you know your product until you put it into the hands of users and

Gladys Mercier: Yeah.

David Miller: once it was in the hands of users then I became truly enlightened and it was really phenomenal to see how they used it as well as to see the friction and the areas uh that can be improved.

David Miller: So, so this this past spring was a really wonderful learning experience um to put the product in the hands of grad grad school students at Carnegie Mellon taking my class on AI engineering and to see and to use a product that is that is kind of like fits fits that mold perfectly.

Gladys Mercier: Yes. Fantastic. So, what were some of the sentiments that the students had? What did you see that they gained from this approach to

David Miller: Well, uh, as you can imagine,

Gladys Mercier: learning?

David Miller: every week when I came into class, I was very eager to get feedback and but I but I tried not to be too biased about, uh, my product and I so I always started saying, "Hey, how was your week? You know, what what did you learn? What was exciting? What was frustrating?" Um and and if if nobody commented on scholar sack then of course I would bring it up and um one one of the the the feedback that I think I enjoyed the most was I said hey what about scholar what's the feedback and and one hand went up and I said yes what what are your thoughts and she said I did not like it I did not like it I was very upset and I said interesting why why was the problem She said, "When I finished the practice exercise, it stopped and I wanted to continue. I wanted more." And I said,

David Miller: "Tell me more." She's like the practice exercise was a guided job

Gladys Mercier: Yeah.

David Miller: interview that is uh centered on the the topics of the week and students um you know get an opportunity to showcase their knowledge in a job interview style format and and it's intentionally set up that way to to be interesting to the students but but also to allow them to self-discover their own gaps of knowledge. Um and and when she had and these these journeys have they're

Gladys Mercier: Mhm.

David Miller: very distinct because they don't go on forever. It's not like a regular chatbot experience. Uh it's very intentionally designed and and it's got an end point and it's designed to be finished in 20 or 30 minutes. Um and and there's a journey that the students go through on the interview and she was enjoying it so much that when it was over she was upset. she wanted to keep going and that is the sort of feedback you want to hear as a teacher. Um and and that really warmed my heart to hear

Gladys Mercier: That's great. That's fantastic. It sounds like faculty also are going to benefit from having a tool like this to give students

David Miller: that.

Gladys Mercier: a little bit more time with the materials but not take so much of your individual time having those one-on ones and the office hours and things like that to help students really get deeper into a topic.

David Miller: Yeah,

Gladys Mercier: Would that be

David Miller: I mean you you you you touch on a really important topic there and one of the things that

Gladys Mercier: true?

David Miller: we do at School is we proudly design this platform to be rooted in educational science and as a teaching community we've got over a hundred years of educational science research that we can lean on. Um the challenge there is students don't spend a lot of time understanding the educational research. So students often don't learn how to learn in an optimal fashion.

Gladys Mercier: Mhm.

David Miller: And actually many teachers don't have that opportunity to learn how to learn. And so one of the things that we do is we take those learning practices established learning science and we build it into the platform.

David Miller: And uh and this really helps the students be more efficient in their learning and it also helps the teachers teach. And so to your point around individualized attention, the learning science shows that individualized attention is the gold standard of of learning and

Gladys Mercier: Mhm.

David Miller: that is showcased in apprenticeship programs that often revolve around trade schools. Also, it's showcased in residency programs that's uh involved around physician training. The problem with that is it's not scalable.

Gladys Mercier: Mhm.

David Miller: It's very expensive and it's not scalable. And through the use of the scholar platform and with the power of

Gladys Mercier: Right.

David Miller: AI, teachers and professors have the ability to intentionally design learning experiences that are grounded in our course materials that have a personality and a characteristic um that that they instill. And these chat experiences as well as voice experiences that we enable uh can be really rooted in the learning science can be grounded in the course materials and can be designed to be an extension of the teacher in the way that that teacher likes to teach.

David Miller: And so we really um think a lot around how can we uh help the teacher? How can we make sure the teacher can use this platform in a frictionless fashion to

Gladys Mercier: Amazing.

David Miller: build experiences that the students can then enjoy in a one-on-one experience with the chat

Gladys Mercier: So this is not just AI applied to learning sort of wrapped around learning.

David Miller: agent

Gladys Mercier: It's actually designed with AI as the tool and learning science as the advice of how to create that product to create those interactions uh for for students learning. How about how about the um the information that the faculty get? How do they know students are actually really learning more

David Miller: Yeah. So, uh,

Gladys Mercier: deeply?

David Miller: one of the great things about the product is it gives faculty direct insight into student behavior as well as student mindset. So faculty have the opportunity through the observability uh uh functionality in the platform to see what's happening but not only do they get to see what's happening they get analysis of what's happening.

David Miller: So analysis you know so if you have a course of 400 folks you might not have time to review 400 chat experiences. However the platform has that capability. platform can read all these chat experiences, can flag uh conversations where students might be spinning their wheels or not progressing at a certain uh pace. It can also detect sentiment across the entire class and surface that back to the professor. Um and you know this was something that we were concerned about in the first course. I I I wondered would students think of this as an invasion of privacy? would they feel reluctant to use the platform because of this?

Gladys Mercier: Okay.

David Miller: And so I had a very frank and open discussion with the students at the beginning of the course and they told me that they felt okay with it. And I thought, okay, maybe they're just saying that to me. And so, um, we did course surveys, a midcourse survey and and a postcourse survey. And we asked questions like this through our survey, which was anonymous feedback to really make sure that we understand the the the pulse of the student and understand what they were saying and what they were thinking.

David Miller: And we got consistent feedback both in person and through these surveys that they felt totally fine. And these students are very tech-savvy. This is you know these these students were the earliest of early adopters. All students they are using uh AI they understand AI they also understand privacy they many of them have grown up with social media. Um and the the private chats that these students want to be private that's happening on their own personal accounts. that's not they're not doing private chats within you know an academic platform and and and they know that. Um I will say in addition the platform is structured to have guardrails uh that that restrict chat outside of class material. So even if a student tried to have a chat around sports or movies or current events, they cannot have that in my course which is centered on AI engineering. So so you know there's a few kind of takeaways. Number one, the students get it and they're okay with it. Um and um the other is the guard rails that we can build.

Gladys Mercier: Mhm.

David Miller: And and and I will say the last thing is you know I based on my observability it gave me the opportunity to engage students intelligently and deliberate uh deliberately around certain concepts where I could see they had interest or confusion.

Gladys Mercier: All

David Miller: If I noted a few uh students having confusion with a topic then I could address that at the at the

Gladys Mercier: right.

David Miller: beginning of my next class. Um, and the other thing that I asked a couple times when I would see really deep, thoughtful questions from students,

Gladys Mercier: Mhm.

David Miller: I would I would say, "Hey, can I use this question in class? I'm not going to attribute it to you, but can I announce your question in class and share it as a wonderful question?"

Gladys Mercier: Mhm.

David Miller: Um, and I always had feedback that gave me permission to do that. So, that was another really cool thing is um to take the best questions that I I saw from a student population and share that back with the class at the beginning of each

Gladys Mercier: Excellent. Excellent.

David Miller: session.

Gladys Mercier: So the kind of analytics and data that you were uh talking about that the faculty have access to is this uh reports or is there a dashboard or is it you know

David Miller: Yeah. Yeah. It's a dashboard. So um you you can imagine just like any corporate dashboard you can see the

Gladys Mercier: visuals

David Miller: number of users. So we've got you know uh activity in the past 24 hours. We can see uh you know which students are using it the most. We can see uh what what students are reading. We can see how they're interacting. We can see it'll tell you if it's voice uh feedback or if they're typing into the platform. You can see hours of the day that folks are using it. Um each session each student session gets a session summary. So if you want to dive into a session summary,

Gladys Mercier: All

David Miller: you can do that. And like I said earlier, there's also this kind of meta analysis where it'll look across all the students and um attempt to uh highlight certain student activity that that demands or deserves attention.

David Miller: And um so there's a lot of great tools in terms of dashboards uh for the for

Gladys Mercier: right.

David Miller: the professor to really uh take a look

Gladys Mercier: Great. Now, how how about how about administrative uh interest?

David Miller: at.

Gladys Mercier: Are college administrators interested in this? Are you know department heads or you know what do they think about a tool like this in the

David Miller: Yeah, great question.

Gladys Mercier: classroom?

David Miller: Uh and this goes back to what I touched on earlier in that we think about four user personas in our platform. We think about the professor or the teacher and the student. Uh we think about the technology team. So often that's an IT group. And we think about administration. administration could be a program director, could be a department head, could be uh the head of a college within a university, or it could be the university administration. And for those folks, they have macro views into what's happening inside of the platform. So they can they have a way to assess are my institutional or college level or department level learning objectives being carried out at the lower levels.

David Miller: So if it's a college uh wondering this they can ask the question is it being carried out at the department

Gladys Mercier: Okay.

David Miller: level is it be carrying out being carried out at the course level and they can basically peer down into what's happening inside of a university uh through the scholar stack platform. So it's a very powerful tool for administration. Um and it it really uh surfaces some insights that are difficult to find otherwise or I wouldn't say impossible to find but

Gladys Mercier: Mhm. Mhm.

David Miller: close to impossible. Insights can be surfaced that otherwise would be very challenging to get answers

Gladys Mercier: Mhm.

David Miller: to.

Gladys Mercier: Right. Right. You often have a lot of manual effort that goes into for example accreditation initiatives. It requires coordination and accurate data collection across many many faculty, different courses, different assignments. It's very hard to kind of prove that you're really delivering on those objectives and uh accreditation bodies are beginning to require evidence of that. So, it sounds like Scholar Stack would be a great answer to help them collect that

David Miller: That's exactly right. So, so we understand all of the different accreditation bodies.

Gladys Mercier: evidence.

David Miller: We understand how they do their work. We understand the materials that they want to look at. And we help the university prepare for those accreditations any day of the week, any month of the year, anytime they want to, they can uh focus on that. And you're right, it's a lot of busy work to to to do that preparation. Nobody inside of a university enjoys that job. Uh and that's the beauty of AI when it's deployed properly. You know, computers can do routine things millions of times and not complain. Um so so these mundane tasks that would ultimately fall on somebody or a group of

Gladys Mercier: Wait.

David Miller: people inside of a university can easily be performed by an

Gladys Mercier: Mhm.

David Miller: AI powered agent given the proper instructions. And that's exactly what we

Gladys Mercier: That's great. That's fantastic.

David Miller: do.

Gladys Mercier: It it covers the gamut from students, faculty, administrators, and the tech team.

Gladys Mercier: You said the uh the IT team.

David Miller: Sure.

Gladys Mercier: What are you seeing uh is their reaction to these kinds of tools being brought into their ecosystem?

David Miller: So beyond the normal questions that any software vendor would get from an IT organization which would normally have to do with security and things like that, some of the unique considerations for an AI tool are well which LLM are you

Gladys Mercier: Mhm.

David Miller: using? Um and at Scholar Stack we use a collection of LLMs. we um can can work with a university or a school that has

Gladys Mercier: Oh.

David Miller: no existing relationship with any LLM provider. We can be that provider and we can bring multiple LLMs to your school and then allow you to curate and choose the use cases for which you'll make them available to your faculty, staff and students. uh for other universities who are more mature might have their own direct

Gladys Mercier: Mhm.

David Miller: contractual relationships with LLM providers and in that case we can easily plug

Gladys Mercier: Mhm.

David Miller: into whatever LLM is being used by the university whether that's Claude or Gemini or ChatGpt if a school has a deal a direct deal with one of those providers um we basically plug into their API API um into their API with the proper access tokens and keys and we can consume and utilize tokens through their direct relationship with their provider.

David Miller: Um, a lot of schools, um, you know, prefer that because they've already built these relationships. They've already spent that money. And so what we do at scholarship is we help those schools get more ROI

Gladys Mercier: Mhm.

David Miller: from that investment. So if you're already paying millions of dollars to chat GPT or to Gemini or to

Gladys Mercier: Mhm.

David Miller: Claude, your goal is to optimize your ROI on that investment. and and that's where we come in is we help uh schools deliver return on investment on the tokens that they're buying from these platforms. And then the last thing I'll mention is another big concern for schools is privacy and and do not train restrictions when you're working with an LLM provider. So most schools who have contracts with LLMs, they have already negotiated uh that privacy and the do not train uh stipulations into their contract. Um and so so they feel very comfortable if we use their um you know their

Gladys Mercier: Mhm.

David Miller: tokens through their provider. I will note that most of the LM providers at this stage have standardized on educational product offerings and you don't have to um specifically negotiate those terms anymore.

David Miller: the enough schools have negotiated those terms with the LLM provider such that there is a now standard product that can be bought that is the the educational product that has most of the same terms that any school would really want from an LLM provider. And that's what we do is when we contract with with our providers, we make sure all of our pipes or all of our agreements are on the educational agreements. Uh and and they're all somewhat standardized these

Gladys Mercier: Excellent.

David Miller: days.

Gladys Mercier: So the last question uh this will be uh an opportunity for you to summarize in uh 30 seconds. What do you think colleges, universities need to do to better prepare students for these new AI saturated workplaces that we're going to see in the future?

David Miller: Well, that is a wonderful question.

Gladys Mercier: 30 seconds.

David Miller: That should be a full another conversation uh because there's a lot there.

Gladys Mercier: Of course,

David Miller: But I would say what I'm most excited about is kind of where I started and that is harnessing all of the power of AI to enhance the learning experience for students uh as well as for teachers.

David Miller: And that is really the main goal of Scholar Stack. We all know AI is incredibly powerful. And we also know that it's not it wasn't uh custom built for educational needs. And so what we do at Scholar Stack is we harness the power of AI,

Gladys Mercier: Mhm.

David Miller: but we deliver it in an intentional manner that is customed designed and custom built for education. And I'm super excited about the future, uh, about all of the possibilities, um, and all of the kind of creativity that it's going to unlock in the

Gladys Mercier: Yes. Excellent.

David Miller: classroom.

Gladys Mercier: Well said. I too am excited about Scholar Stack. Thank you for being our guest today, David Miller, founder of Scholar Stack.

David Miller: Thank you,

Gladys Mercier: Check it out.