Nov 20, 2019

Ray and Dan are joined by Dr David Kellerman, who is a senior lecturer in the School of Mechanical and Manufacturing Engineering at University of New South Wales (UNSW) Sydney. 

David believes that the way to engage a growing student body is not to reinvent pedagogy, but to create a version of it that works at scale - one that produces, from a disparate and growing number of individual students, a learning community that becomes smarter from technology, not displaced by it. Woven throughout David's work is the use of Artificial Intelligence to personalise learning, support students and create a truly engaged learning community. 

In this episode, Ray and Dan get the chance to hear more of the journey that David has been taking his students on.

TRANSCRIPT FOR The AI in Education Podcast
Series: 1
Episode: 9

This transcript and summary are auto-generated. If you spot any important errors, do feel free to email the podcast hosts for corrections.

This podcast episode features an interview with Dr. David Kellerman from UNSW Sydney, focusing on his innovative use of artificial intelligence to personalise learning for large engineering classes. Dr. Kellerman addresses the challenge of scaling a humanistic educational experience—where an instructor can provide genuine care and individual attention—to hundreds of students. He explains that his solution involves creating integrated systems and platforms, leveraging modern tools like Microsoft Teams, OneNote, and Azure cloud services to overcome the inefficiency of disparate educational software. By digitising all activities and bringing data onto a single, governed platform, he developed a system that uses machine learning to accurately predict student performance on individual exam questions, allowing the automatic generation of personalised, competency-ranked study packs to ensure comprehensive mastery of the curriculum. Ultimately, Dr. Kellerman aims to open source this technology built on accessible platforms like Office 365, advocating for quality education at a global scale.

Hello. Don't panic. You're definitely listening to the artificial intelligence in education podcast. I am sure you remember my voice from the alien accessibility episode. This is a special feature length episode. So before you hear any human voices, I am here to fill you. about the background to the interview. In July 2019, Dr. David Kellerman from the engineering faculty at UNSW Sydney spoke at Microsoft's Inspy conference. 25,000 people packed into a stadium to hear him explain how he was using artificial intelligence in Microsoft Teams to personalize learning for classes with more than 500 students in them. In this podcast, Rafe and Dan interview David, and I will interrupt to bring you some excerpts from his July conference speech. Let me play you David's beginning from the conference and then I'll let Den and Ray into the recording studio.
So, my name's Dave. Good day everyone. Today I'm going to talk to you about teamwork and collaboration with 500 students on and off campus. So, I come from UNSW Sydney, a big Australian university. We have 65,000 students. Our engineering faculty has 17,000 students. That makes just our engineering school bigger than the whole of Stanford University. And we've got classes of 500 plus. But that's 500 islands, right? 500 people looking down, taking down notes from a document projector, consuming PDFs, watching canned content. So, here's a challenge. How do you get 500 students to work together as a single team, as a learning community, whether they're on campus or off campus?
That is David sharing his story at the Inspire conference. So, as I promised, I'll let Ray and Dan back into the studio now.
Well, Dan, we're here again.
Yay.
So, more about artificial intelligence in education.
Fantastic.
Hey, did I tell you AI? We talk about AI all the time. I was talking to a principal in a school yesterday and they were talking, they're in rural Victoria and they said AI to my students who all come from farming stock that means artificial insemination. So, I think we need to be clear. We're talking about artificial intelligence in education.
Yeah.
Yes.
Okay. That could be for the cutting room floor, couldn't it, Dan? Okay. So, uh Dan, do you want to introduce our guest today?
Yes. Today we've got David Kellerman from the University of New South Wales who's done some fantastic integrative projects and use AI to really boost the outcomes of the students. So, we're really going to try to bring the story to life and really investigate where AI is going and how you find the impact across your classes and where you think the future is.
Great. Thank you. Dan, thank you David.
Thank you very much. Yes, I'm very glad to be here.
What have you done?
What have I done? I've done a lot of things, but there is a a basic guiding principle to what I'm doing, which is I think that all of the innovation space in education is to do with integrated systems. Systems that leverage one another. And if you look at the pedagogical landscape or the landscape of educational software, it's all about out individual apps. The learning management system itself is a self-p profofessed hub and spoke system. It's lots of different things that plug together and they don't really integrate very well at all. So, what I have been doing is integrating lots and lots and lots of different components into very very modern software platforms and then leveraged a number of very modern services like cognitive services bot framework, cloud services, uh simultaneous document editing, automatic transcription services, all of these different kinds of things to leverage that integrated platform.
So what was the initial problem then? You mentioned there's lots of these disparate systems, but in your context, you know, you've got lots you're teaching lots and lots of students.
Um so what what were the problems that were occurring? What were you trying to solve?
Well, so to give you an example, I'm teaching two classes this seme My bigger class is 600 and my small class is 500.
Okay.
So I'm teaching,00 students. I'm the sole course convenor for those courses. And so I am responsible for an incredible number of student experiences. And I like to be as personable as possible and as available to all of those people.
Yeah.
So my real challenge here is how do you scale an educational experience that maintains the kind of humanistic level. And you know, for anyone listening here, think back to maybe when you were in year five at primary school and there were 25 kids, say in the classroom and there was your teacher, Mr. or Mrs. whoever, and they knew everybody by name and they checked on you and they said, "Oh, where's where's little Johnny today?" Or, you know, they walked out in the playground and they made sure that somebody wasn't being bullied or they gave extra work to Maria who was great at reading and actually they were delivering and they still do to this day an incredibly humanistic experience. Now when I use that word it is often misconstrued in different ideologies but there is a particular ideology which is humanistic education and it is educational delivery that's driven around personal needs empathy and these other things. So when we take an educational experience like that and we scale it to some horrific scenario where one poor professor is in charge of 1,100 people. Then you say, how on earth can that one person be spread out so incredibly thin as to be able to maintain that level of empathy and genuine care and connection to that many students.
Now it might seem impossible, but it's not impossible. We can can do it and it requires a lot of thought and in my case a lot of engineering
and I am a mechanical engineer and the goal which is what I tell my first year engineering students is that we have a very very privileged role in society which is our job is to make the world a better place. Our job is to make airplanes that are more efficient and smartphones that are smarter and cars that use less fuel and houses that are more comfortable and thermally efficient. Our goal is to make things better. And so I just look at these systems from an engineering point of view. I say where are the inefficiencies? Where are the places we could make this better? What are the places we could save times doing things that don't matter so that we can spend more time doing things that do matter?
And and that's true in education generally, isn't it? Because when we're looking at the impact of what actually really makes a difference, it's the lecturer, it's the teacher, it's the person at the chalk face, and there's a lot of administration and there's things in the background. around lots of assessments and things like that which are important but sometimes they do take away from some of those efficiencies. So, so when we start to kind of unpack this what would it actually do simply? Okay. Yeah, great question. Obviously, I've got a I've got my own sort of organic story and it takes a little bit too long to tell that. But to give you the really short version of it, I I started teaching a class with 550 students a few years ago and And the only place we could fit them was in the graduation hall. And there was a podium at the front with a document scanner, which is basically a camera that points at a sheet of paper. And the students would watch a big giant 50-ft projection of the back of my hand, basically writing stuff down. And then at the end, they would a few of them would put up their hand and they'd say, "Oh, sir, could you scan that and upload it to the learning management system?" And I walked out of that room having used the default systems and delivered the course in the default way and I just thought this is ridiculous. So the first thing I started doing was using digital ink on a digital ink device really great quality and you know you don't have to scan anything cuz it's already digital. Can I ask a question on that one? So I did math teaching myself you know in a previous life. What what are your thoughts on that kind of entire process? Because obviously I think one of the things with technology there's been this gap previously with engineering and maths because you know everybody will go well. It's easy to teach English or certain other topics using technology but maths and engineering has been quite tricky because you've got to put specific mathematical formula in the things.
Yes. So engineering I think is probably the hardest there is.
Yeah.
And the reason is that we've got all of the difficulty of communication that mathematics has which is complex mathematical notation glyphs and so on but we're also in the laboratory. So you go to medicine and you say, "Well, they're in the lab looking at a kadaavver or whatever." And you look at maths and you say, "Well, they're writing all of these complex mathematics
and we're doing both of these things at the same time." And of course, the same argument goes for science much of the time as well.
So just using that technology on it own is that did that have a big impact
for yourself?
Yeah. So digital link was hugely valuable and of course the next step in that story was what if I didn't even have to share that digital linking to the learning management system. What if the students actually owned that same notebook as well and we were simultaneous collaborators on that notebook?
I started doing that in 2017 with a tool that came out called class notebook which is an addin for oneote and what that basically means is that all of my 500 students became members of that notebook. When I wrote an equation it real time synchronized to their computers. They could have a tablet open in the lecture and my writing would be appearing on their screen. Now, the interesting thing about that is a lot of people looked at it excitedly and said, "Oh, you know, we've got this ink to maths conversion and you can turn this into equations." And I've had this said to me so many times by excited people. And I'm like, "No, no, no.
I don't I don't want to convert this. This is a
really good point. This is a
pivot." Yeah.
It's like telling an artist to paint a picture of a horse and then go, you know, we can recognize your painting and then give you a perfect picture of a horse instead. Yeah. No. Okay. This point this writing is is a very very authentic thing.
And this is one of those very beginnings of coming back to that very humanistic experience which is that writing something that we did together that I was at the front asking students help me with this equation. They saw everything been drawn and inked together. We did it together. It was appearing on their screen as we did it. This is a very authentic emotive kind of thing. They they have a real visceral connection to it. So we want to hold on to that and now we want to take that as a launching point to say how do we make every single part of this educational experience be something authentic that was done collaboratively, cooperatively that we created together. We didn't
log into a learning management system and hit play video download PDF. open to page 37. We create the material together. It's very very authentic.
Can I ask a kind of boring question under there as well? You know, in in universities, you know, there's a big behemoth of a kind of system where you've got to bring curriculum out and it's quite
I suppose from the outside sometimes looks hard to innovate. Yes. Because people will go well we're using X learning management system everywhere. This is what we've invested in. We've got lots of people that monitor quality of curriculum courses and things like that. Did you manage to push the boundaries there when you've got the the quality of learning and teaching team or whoever it may be, you know?
So, so it is hard to innovate when you're in a silo. Okay. So, let's just imagine for example, you're a chef and uh I slide over to you, I don't know, a bowl of potatoes and I say, "Inovate, right?" And you go, "You know what? No one's ever no one's ever cut the potato into trying angular pyramids and baked them before. We did chips, we did crisps, we did all of these. I've got a new shape and you bake it and try and sell it off in your booth at an educational technology conference. Um, okay. So, scenario two. What if you open the refrigerator to a chef and you say, "Here's a whole bunch of different ingredients and I would like you to create a new dish." Well, now they've got a chance.
So, this is the way we have to Look at it. Yep. If you say to somebody, "Hey, here's some little bit of software that students can use for writing their assignment, innovate.
Everything's been done, but they not but there isn't much scope to actually do stuff because it's in a silo."
So, so I'm going to take your first example of I'm going to make teaching in the classroom, the lecture hall more authentic.
Yes.
Digitally. That's the That's the new style of potatoes thing. because it was it's still 500 people in the lecture hall watching you do something and there's more connection and yeah
yeah so then so then the next journey is that innovation beyond the box accepting there are some rules in the organization to follow along but it's the how do you make a more personal experience
all right so I'm going to extend my chef in the kitchen analogy okay so you go what you know what though it's we don't just have potatoes we've got all of the different ingredients but here's the problem every one of the ingredients is in a different kitchen And you're not allowed to take it out of that kitchen.
Okay?
So you can you can have the potatoes in kitchen A and the butter in kitchen B and so on. So you got to get everything into the same platform. You have to interconnect things. So what I basically looked at were two main arguments. Number one, you need to digitize everything. The digital ink is an example of a native digital experience. Now yes, when We write on paper and we scan, we are digitizing as well and that's perfectly valid. But when we have the opportunity to do native ink, then we have a native digital experience. Part two is how do we get everything on the same platform. Now I'll use our own institution as an example. Learning management system is on Moodle. Lecture recording system is on Echo 360. Word and email is over on Office. the students log in or enroll in their courses on u unsw.edu.auu. All of these different experiences are on different platforms one after the other after the other.
So now you say okay well lecture recording that was one I just mentioned. Firstly we've got to say well what platform are we going to choose? Now the obvious thing is well you choose your learning management system don't you? Isn't that the thing that talks to everything. But the problem with learning management systems is that we don't actually do anything on learning management systems. They just tell us what to do somewhere else. So for example, write your essay. Nobody ever puts the cursor into the learning management system and starts typing. They go and they open up
Yeah.
Word, Latte, Google Docs, I don't know, whatever, and they start doing it somewhere else.
Yeah.
So let's start making the hub the center of where we're going to integrate the place where we actually do stuff. Now lucky for me the place where I started working on this inking which is one note is the same place where most of my students write their lab reports which is word whether they use a Mac or a Chromebook or whatever it is at university most students are using word or latte.
Okay. So now we start looking at a common platform and at the back end of that platform we've got a little of really powerful backend systems. So for example, cloud storage, where does the file live? Are we downloading the file and then uploading the file or does the file live in the cloud? And as soon as it lives in the cloud, what about simultaneous document editing? If we're talking about collaboration, then our documents can't live on hard drives. They have to live on the cloud and multiple people need to be able to open them and access them at the same time. We've got to get rid of the cycle of download up upload, go to an external service, come back to where you started. So, of course, we have one drive for business at UNSW, sharepoint uh in office 365, Azure cloud services. All of these systems are sitting at the back end of these doing platforms like word and excel and oneote and so on.
So, that's really interesting because in a lot of the conversations around using AI, one of the first question questions people have is well where's our data and they have to go and spend time pulling their data into one place in order to be able to use it for predictive analytics or whatever.
Yeah,
you did the other route which is bring all of our activities into one place.
Yes.
And so when you went to the next stage it meant your data was already there.
Okay. Yes. And this is the pertinent question and this is where my engineering hat is on very very firmly.
I can confirm that I can't see you
which is not only are we on the platform where the doing is but like I just said all of these backend services exist Azure and SharePoint and so on and what I then start doing is saying that all of the data surrounding these activities is maintained and stored on that same platform and so in this case we're talking about Azure SQL or a data warehouse or a database or whatever it is that data is generated on that platform and that data is stored on that platform. Now that means that we put on our other hat here for a sec which is our data governance hat. We have a system that is GDPR compliant. That's you know general data protection regulation that we have a system that is end to end encrypted that is on a platform with a strong form of authentication this is AAD and has data sovereignty which means it runs on servers which are in country. So that's Azure data services in CRA. So we've got very very tight acquisition of data. We've got very very secure storage of data and we've got high levels of accessibility of that data to then do stuff with it. And when I say do stuff, well, you aggregate data from these different systems, but at your fingertips, you also have machine learning, you have bot framework, you have all of these other basically what we loosely call AI systems. on the same platform where they are also generated and stored. So now you have a closed loop system.
So if I'm um if I'm an advertiser then that that world to me is great now I can I can service really personal adverts or if I'm a online reseller I can say great I can sell them the next pair of shoes I want them to buy.
So what do you do with that from an education perspective?
So the other thing is that well firstly universities we are not profit organizations. There there's no such thing as profit at universities. Uh any money that there is goes straight back into the research and the teaching infrastructure of the university. And the only single goal that I have is to generate a better student experience. That's what we do, right? Universities do research and education. And if we're not doing those two things, we shouldn't be here. So the only question that we ask with these systems is how are we delivering a better educational experience with them. So in the case of this data when our students enroll and most universities will have this it will say that the university has essentially an access and a right to use data so long as it is used for the purpose of educational delivery.
I'll interrupt again because I want you to hear how David described this work to the attendees at the conference.
So lastly I want to talk about closing the loop. We've leveraged this value that we have. We've leveraged all of these digital assets and we've integrated our solution. But how can we turn all of that data into new material? In fact, by integrating all of these systems, I have very highly structured data on SQL. So, I used 2017 data to train an Azure machine learning model to correlate all of the information against student performance and using a database of competency ranked resources. That's every every single topic against every difficulty level and an algorithm onnet. We automatically assembled 500 individual personalized optimized study packs for every student based on the prediction of not only their exam result but their exam result for every individual question two weeks before sitting it and uploaded to SharePoint with personal access oneclick in teams. Pretty cool, right? Okay, that's it for me. Let's go back to the studio where Ray and Dan are listening to David talk about how they're personalized for students.
And that is exactly what I do. We take the information about the students and we use it to give them a better educational experience directly.
This is where we start getting into terms like tailored learning and personalized learning experiences,
which is quite big in in K to2 as well. You know, there's a big push at the minute in Australia specifically around that personalization and and schools find it very difficult to do. So
well so let me dig a bit further because it's a buzz word personalized learning buzz phrase.
Yeah.
Um everyone talks about it but what does it mean to your students specifically? What do they see that helps them to understand what it means?
So this is one of the fantastic challenges of higher education which is we've got a whole bunch of these buzz phrases
and every university has the same one the same ones. personalized learning, student retention, uh, learning communities, all of these different
K12 as well.
We use all the same terms. Secondly, K to2, oh yeah, they're light years ahead of us. Remember Maria who was really good at reading in year five and got given the extra reading by her teacher? That was a personalized learning experience and it was fantastic and her education was better for her.
The question is is how can we bring that to scale and The question that I always ask when I look at these uh buzz phrases is, hey, how do you actually do that? How do you actually deliver it?
Because the reason why these phrases have become so dominant is because they're kind of like grand challenges for universities.
They're stuff that we all aspire to doing. They're all on our mission statements, but nobody really exactly knows how to do it.
It's very true. Yeah.
So, we can take an example. like a personalized learning experience. There are universities where they are touting that they're using AI to identify students at risk and you know then they're intervening but what they're actually doing is that they're basically taking a file of how often one of their students logged into their learning management system
and they're training it against the grades of the students. and they're saying, "If you don't log in very often, we're going to automatically send you an email saying we're worried that you're not going to pass."
Okay, now this is, you know, and I apologize to anyone who's listening who that is their system, but it is comically bad because there is absolutely not even close to enough information about that to understand people. And you have to have a huge amount of information, especially for machine learning. to start to get to the point where it is actually very very accurate and we can get there. Everybody has had the freaky occurrence where an ad predicts something that they were just thinking about and that's actually done with machine learning. We can get to the point where we have these incredibly accurate models of where people are at. But you need enough information and enough data.
So you gathered that information in from the various systems in inside your governed cloud. and the various technologies have brought that together. So what did you do with that information then? So where where did it start to go from you know you were doing this as a kind of innovative educator to then starting to say well hey you know this is how far I could get with this now where do I need to get say expert help from outside where can I you know where did that boundary start to
okay delineate so uh data prediction action I'll go through these really quick so How do we get the data? Well, we need a data schema.
Now, a data schema is something where we can't you can get a data lake, right? You can just get heaps and heaps and heaps of random data and just throw it all into a big pool and then ask ML to do something with it and it's not going to be great. If you've got some kind of structure that you can add and that structure is a data schema, think about it like you've got a whole bunch of objects and at the very least you're going to throw vegetables in the vegetable basket and you know this kind of thing. I have a data schema. My syllabus is broken down exactly into modules that map against each week of the semester. That's really important because we live our lives on a weekly cadence. It's the weekend on Saturday and Sunday and we go to work on Monday and we have a lecture on a Tuesday and we've got a tutorial on a Thursday.
And so we want to break our syllabus into weekly modules. Then uh I use teams which is also on office 365 for communication. I have a channel for every single one of those weeks. We're going to talk about what's going on that week. We're going to put the resources that are required for that week. So that's a lecture recording, the notes for that week, the handwritten, the digital ink that we used, the discussion surrounding that week, the tutorial problems we're working on that week. They're all in that bucket. But then at the same time, we can have uh quizzes that are to do with those topics. And that data can go in the bucket. And did they watch the online recording? Did they go to the lecture, how did they do in the tutorial that week? Every single thing can fall into that bucket. Now, that generates something called a key value pair. And a key value pair is a strong correlation between two things where we can tell an ML process, which is basically just training a neural network. And a neural network is just it's a universal function approximator, right? So, you've got an input and an output, a whole bunch of hidden layers, and you're basically training different waiting functions to get a mapping between this. So, when you create these key value pairs, you get a much stronger a threat of waiting functions when you're training that machine learning.
So now you have these strong correlations like let's say Ry went to every lecture except week five and the quiz to do with week five he did badly.
It's always me I've noticed the example
or alternatively. So so when you have that that very very strong mapping between the the quiz in week five was to do with the attendance in week five compared to Ray went to n out of 10 lectures and he did poorly on one out of the nine of 10 lectures.
Yes,
it's a lot harder to confidently map those two things when you don't already are priority know the correlation between those two things.
Now remember that everything we're doing is authentic and being generated by us. We're not consuming content. We're making the live recording. We're making the ink. We're having the conversations. But we can do this within a framework that is prescribed our priority. And that framework is of course the syllabus and the calendar. We're going to map those two things. We're going to do what we do very naturally and organically. But everything that happens is going to have telemetry and outcomes that's going to fall in between those specific buckets, generate key value pairs, and allow us to generate very, very accurate ML models. So you take of this data and then I have an exam and the exam has got the exact same number of questions and topics as the semester and the weeks and that exam I have built a special system where we scan those exams and we mark them online and the reason is that we've got all of the exact data for every individual component of mark
so and is so those exams then are done you know standard exams on a piece of paper in exam hall
you scan it in and then using a a sorry It brings it all back together and
yeah and we use a bunch of AI in order to digitize to read the name of the student and everything but it's still human marked
and now you've got a data set where you take all of the input and the exam is the output and you train an ML model. The ML model basically generates a whole bunch of waiting functions and structures. You I lift that model from the previous time that I ran the course. In this case the first time I got that training set was 2017. I apply it to the class in 2018. 18, for example, but I do it as soon as I've collected the data, but before the exam. And that allows me to very, very accurately predict the performance of every one of my individual students for every individual exam question a couple of weeks before they've sat the exam.
So now that's the prediction. So we've got the data. Yeah.
And then we've got the training. And then we've got the prediction. Now the last part that we have is how do we now return value? Because that's the only goal that we have. How do we return student value uh experience and so on. So what I have is I've got a large pool of resources that were painstakingly created which is essentially uh work for those students to do on every single topic at every single competency level. Uh now if we have somebody who let's say Dan who is going to get an HD and we think that he's done absolutely brilliantly in the week five topic well I'm going to give you some really, really challenging topics because I want to, you know, bring out the the potential you have here. I want to actually push you a bit, go, not go, "Ah, that's an easy topic."
Now, Ry on the other hand, who who missed the week five lecture and was totally flunking.
What is Ry going to do when he comes up to study for this exam? Well, he's probably going to have some narrative to himself like, I really need to pass this course, that kinematics topic we did in week week five.
I definitely didn't understand that. kinematics.
I miss that. I ask people and I still don't know.
And you still don't know. And you go, you know what? I'm just forget it. I'm I'm not even going to try and answer that question. I'm going to focus on the other topics and I'm going to try and make sure I pass. I can do those questions. I'm going to brush up on them and I'm going to just sacrifice. That's basic strategy, right? It's like the way armies go, well, we're definitely losing that battle. Let's pull the troops and put them over here. This is the way we think now as a university, especially within an engineering degree that is accredited that wants to deliver people with competency in their profession. I want to deliver students who are competent across the entire range of learning outcomes and topics. So we use these predictions and I algorithmically using net produce a personal study pack for every single one of those students. It gets uploaded to SharePoint with individual access rights. Then there's a dashboard that shows a rich display of their performance. and one click, here's your personalized study pack. They click on it and there it is. And Rey has a topic for week five
and it's got right from the very beginning walk through problems. Here you go, step by step. We're going to get you to a basic level of competency. You're going to be able to do these problems and you're going to go, you know what, actually, I think I can do this and we've got a chance of bringing you up to this baseline competency of this one.
And because everything's tagged there, Yeah. If he's stuck on something, it can take him to a part of the lecture that you've just done in week five.
Exactly.
Yeah.
And and hugely relieving for somebody who's just about to fly.
The fact that the engineer that's building my plane didn't miss the bit about propulsion.
Yes.
Is quite an important thing. That bit about with most situations you can Yeah. You miss one bit, but you carry most of it. You know, the meal was mostly great, but the potatoes weren't great.
Yeah.
But in the case of engineering things, every aspect is
So that that's a fantastic step through. What What about the just to you know we're coming to the end of this now I suppose but is you you're itching to ask something
I was thinking so we've been really lucky because we've been able to put in here because you've talked about a lot of what you're doing publicly we've been able to put into this podcast some other things you talked about publicly and then relate it you've given us the backstory behind it
so probably the really big question is everybody else listening to this going that's great but that's David how do we get to use some of that skills and expertise how do other people get to leverage the work you've done David Yeah. So, big grand challenge, one of those big buzz terms, quality education at scale.
Mhm.
That is what I'm interested in. I'm interested in how can we bring that great experience that the 25 kids in year 5 had all of the best elements and deliver that same level of care and empathy and individual experience to 600 students instead. Now, that's one question of scale. The other question of scale is how can you take that system and scale it to an entire university. But by the time you've asked that question, you need to start thinking, well, hang on. Why should we stop there? How can we scale it to any university? Now, this is where you look at the strategic platforms. You say, how can you deliver this to the broadest audience? Well, the first thing is let's look at those platforms. So, we have a whole bunch of learning management systems out there. We've got Canvas, Blackboard, Moodle, all of these different platforms, they all have market shares that are not hugely different. You know, they're dotted all around the world.
But you know what? Almost every single university in the world has Office 365. And Office is given to students for free. So kids in schools can go and get an Office 365 account for free and they can start using Word and whatever. So we've got this platform that is hugely powerful that has all of these framework for AI and bot framework and so on that 90% of universities in the world use that's basically free for students that I can build on. That's the best leveraging I can possibly see for being able to work on something that's going to have impact across the global education audience. Now the second thing is well what are we doing? Well I'm building things that work within the development framework of those systems and What I'm trying to do is basically opensource the code of the tools that I built on top of that platform. So fantastic. We I I think maybe another time we could maybe do another podcast or something about conversational agents, but another thing that I have is a a bot that works very strongly on connecting humans with humans rather than humans with bots. I have this tailored learning system. I have all of these various systems that are built upon that framework and I am open sourcing those systems. So That means that there's a GitHub repo that any university can download that codebase. They can modify, they can contribute to the global codebase.
So now what you have is all of this technology that's being built on open- source code that runs on a platform that's at most universities in the world that's even free for a little school in a village in Africa.
And what's the date? What what are you aiming for in terms of having that GitHub out and available for people?
I mean that's just about basically done. That opensource repo is already out there. I don't think
uh we'll probably open that up in the next couple of weeks. But that the the first step of that open source project is basically done.
I've got other universities already who are working on deploying those systems. But I want to go further and further and further with this. I want to start looking at basically the IT platforms that universities use like their student information systems. Uh how they use things like dynamic and CRM and ERP and all of these different systems.
That takes us back to the beginning about AI. I bet there's different definitions of CRM and
Exactly.
Okay. So, that's great. So, we can put in the the show notes the details where people can go and find out a bit more about this.
Yes.
The other thing is I know that Cloud Collective did some work on the coding in the back end of this and they're continuing to do some work on it.
Yes.
So, should we go and have a chat to them as well and find out from a technical point of view where they are?
Yeah, I think that would be a great idea. Um,
it's a good idea for the next podcast, right?
Let's go do that next time.
And and Tres are the group there that I have worked with on a lot of the development and there's some fantastic people there who really took on the challenge of not just sort of delivering what was on a statement of work, but of putting their own experience because of course they've all been educated and they've been to universities as well. And it's really easy to say to someone, imagine an older student, what would you want this experience to be like?
Well, thanks David for opening up the engineering faculty today to us and and sharing your thoughts on AI and the way of using it and we look forward to chat and tune in later podcast.
Thank you so much. Thank you. Thank you Ry.