#498 Edge AI Meets Global Impact: Dev Aditya on Scaling Learning Without the Cloud

In this episode, host Mehmet Gonullu welcomes back Dev Aditya, Co-Founder of OIAI by Otermans Institute, who returns nearly two years after his first appearance to share an extraordinary update: they’ve built an AI teacher that runs completely offline — even on a 2020 Android phone.
From educating learners in refugee camps to scaling across 17 countries, Dev’s mission is clear: make AI-powered learning accessible to the 750 million people underserved by the internet and modern hardware.
We explore the engineering breakthroughs, the philosophical pivots, and the global partnerships that are making it possible.
🔑 Key Takeaways:
• Why offline, voice-first AI is a game changer for education in low-connectivity areas
• How Dev’s team runs 3D avatar-based AI on edge devices from 2020 — without cloud inference
• The reason they pivoted away from enterprise clients to return to their impact-driven mission
• Why running AI for the marginalized means solving for both connectivity and hardware bottlenecks
• Their ambitious goal: upskill 750M people by 2030
• The launch of an SDK for other organizations to embed their AI teaching system
• Upcoming partnerships with African governments and UN bodies
⸻
🎓 What You’ll Learn:
• How to scale AI without relying on cloud infrastructure
• What true digital inclusion looks like — beyond donations and broadband
• The importance of philosophy and mission in product decisions
• The future of immersive, voice-first learning experiences
• How to build for constraint — and still scale globally
👤 About the Guest:
Dev Aditya is the co-founder of OIAI by Otermans Institute. He’s known for launching the world’s first public-facing AI teachers well before ChatGPT entered the scene. Dev’s work focuses on democratizing education through AI and edge computing, bringing scalable, impactful learning to the most underserved populations.
https://www.linkedin.com/in/dev-adityaofficial/
https://www.portfolio.oiedu.co.uk/
00:00 – Welcome back, Dev Aditya & what’s changed since Episode 212
02:00 – From rural schools to refugee camps: The origin story
05:30 – Content vs. teacher: The real challenge in scaling learning
07:00 – The philosophical misstep and going back to the mission
09:40 – Monetizing impact: Lessons from enterprise and education
11:00 – Offline breakthrough: Running AI without internet
13:40 – Engineering feat: 3D avatar AI on Android 2020 phones
15:20 – Why bigger models don’t equal better outcomes
18:00 – Did open-source LLMs influence their vision?
21:00 – How they handle data, analytics & personalization offline
23:00 – No patents? Why they’re releasing an SDK instead
25:00 – The SEA Model: Skills, Efficiency, Accessibility
27:00 – Building voice-first digital experiences for the underserved
30:00 – VR + AI for marginalized learners? Already built
32:00 – Teaching entrepreneurship, not just STEM
34:30 – An AI teacher that remembers you
36:00 – What’s next: Kenya, Guinea-Bissau & São Tomé
38:00 – Final thoughts & how to get in touch with Dev
[00:00:00]
Mehmet Gonullu: Hello, and welcome back to a new episode of the CTO Show with meme today. I'm pleased to welcome back again after almost two years, dev, who is the co-founder of OIAI by Otermans [00:01:00] Institute. Dev. It's so nice to see you again on the show. Like, uh, I feel very happy when I see some of my guests come again and this time to tell us more where they have been in the journey, just as a, you know, reminder for the audience dev.
So more about you, what is OIAI? And then we can start the conversation from there.
Dev: Yes, absolutely ma'am. It's, uh, great to be back here. Um, for us, I think, yes, it's almost been two years, but I think that's, that's a good sizable chunk for us too, and for me to update you, your audience, on what we've been doing and we've been doing a lot.
So for your audience, a little bit of a reminder, uh, you know, a little bit of a sort of update as well on what we do. So the world possibly knows us as the people who launched the first public facing AI teachers. This was in January of 20, 20 18, months before chat GPT. [00:02:00] Right. And since then we've had three versions of our product.
The last time we spoke, I think we, we were touching about 10 countries. A lot of them at the surface now we are servicing learners in 17 countries. Right. But I think this is a very exciting time for us to talk because what we have also now been able to achieve is run our AI completely offline. So perhaps.
For the next months, maybe even years to come, the world would know us as the ones who are doing education really democratically and enabling it to run in places where people are not even using Google on a regular basis because of internet bottlenecks. So good to be back.
Mehmet Gonullu: Exactly. It's good to be back dev And, uh, thank you again.
Now, you mentioned about the story just for the audience who might missed, you know, that episode, the episode number is, by the way, uh, 2 1 2. So it was in September, uh, 2023. Um. The idea itself, dev, like if, [00:03:00] because I'm gonna start from there and then talk about, you know, the latest breakthrough that you have done.
What was the main motivation? And I remember there was a lot of struggle you talked about at that time, what was the motivation? What was the, as they call it, the aha moment for you and, um, you know, to, to to, to go and say, we need to build this. What was the main, uh, use case?
Dev: Absolutely. So, uh, it's always good to touch the, you know, the roots of the journey, right?
So I'm gonna be talking about that. Um, and I think since our last talk, we've further learned a lot ourselves as an organization as well, I think, which is very important. So let's just start with the story, right? So my co-founder and myself, uh, she was president, I was vice president of Brunell University London, and we worked very well together.
So after that we said, let's. Continue our work in education and we decided to go to South Asia, so mainly Nepal, Bangladesh and India. And [00:04:00] to teach some people, right? We are talking about rural schools, madrassa schools. One of the schools, one of the schools we were teaching was inside a national reserve.
On the second day we saw leopard. So real rural, right? Yeah. So that's how it started. Okay. So it went well. We saw real impact on the ground with these people, and then we said, let's. Formalize this. Let's make it into an organization. Okay, so we did that in February of 2020, and within two weeks what happened?
We all remember it. The pandemic happened. Okay? So we had to go digital overnight completely. Now that was good in its own way. Because suddenly in the next couple of months, we were servicing 35,000 people. We were servicing for U-N-H-C-R camps. We were servicing for unicef. We were in about nine countries.
Okay. And one of the things we realized in our journey was there's only so much content. Can do in learning, in the dissemination of learning. So just imagine [00:05:00] this. You're gonna give people free books, you're gonna give them YouTube, Khan Academy if internet allows, there's only so much it can do, and there are a few people who will really persevere through that to really learn, okay?
The gap actually exists in having a teacher, a trainer, a mentor, somebody who can answer your questions, who can encourage and motivate you to learn, who can practice scenarios with you, et cetera. So that's what we were doing. We were human led back then, but when we reached that 35,000 threshold, even with volunteers, et cetera, helping us, we just couldn't scale further because it just doesn't work right?
That that's sort of a model. So that's when we thought, what's a teacher? A teacher is intelligent. So this is way back, you know, before this new cool wave of AI started. But we said, let's try ai. It's got intelligence in it. It's got that automation. Feel to it, right? So let's try it. So we decided to try that and luckily our first product was ready and launched in January of 2021.
So that's about [00:06:00] 18 months before GBT mind you. Right? And our first students were in a refugee camp, U-N-H-C-R, backed BCF backed refugee camp in Northern Iraq. So that was the story. Then. Now, fast forward from then to when we spoke to you the last time. That was version two of our product. All right. So as you can see, at the epicenter of what we do is always about supporting the marginalized, right?
It's about empowering these people. We have a mission to upskill 750 million people by 2030. Okay? But when we last spoke. That was essentially a period of mistake. Let, let me just say that, although I was very excited about it, then one of the initial questions I had at that time from investors, from other people is, okay, you're helping them, but who's gonna pay for it?
Okay? Mm-hmm. And that sort of got us in a wrong direction. So when we spoke to you, we were hitting hard for companies L and d, [00:07:00] right? We were hitting hard for higher education. We were hitting hard for schools, okay? That was not the right thing to do. Today we are in 17 countries. We are still servicing schools.
We are still servicing higher education. Right. We are still doing that, but we, after about a year of what I call, you know, sort of. Uh, loitering in the desert, if you wanna call it that way. We realized that we should go back to our dream because we were thinking we are gonna get some money from these people and then go back there.
Right? It just doesn't work that way. So we had to go back to our dream. So today we are back in our dream and to make this dream, we consciously work for the last year and a half to make our AI offline because in the marginalized space, there are only two bottlenecks. It's hardware, right? And it's connectivity.
So we've solved the connectivity gap. We've also solved the hardware gap because ours is the only system that can run on edge in [00:08:00] 2020 or post style phones. Okay? You don't need the latest Samsung or whatever. So we were able to do this. Now we are deep in places like Africa and Latin America. I'm sure we're gonna talk a lot about it.
And in August we are going to do. Africa's largest and the world's fastest AI literacy upskilling. So very exciting, and it's, it's really refreshing to be back in, in line with our original mission.
Mehmet Gonullu: You know, I remember. One of the thing that stopped me when we talked the last time, and of course we kept the conversation is this, uh, idea of yeah, democratizing learning.
And I remember about also like how the AI as teacher, um, would adapt. To different cultures, you know, because you deal with multiple countries, different cultures, and people in need. And I remember the example you gave last time about like, imagine maybe, I don't know, it's a rural area in [00:09:00] some place, maybe in Asia, Africa, and they don't have access to teachers, and now AI can come here.
So this is by itself fascinating. Mm-hmm. Now, before I go to the technical part, you mentioned about the monetization. Part of it. Mm-hmm. So what have changed from that perspective, dev, like of course, you know, kind of impact projects like what you're doing. And I know for a fact a lot of people would still invest in this, but they would ask again, like, how would you commercialize this?
Um. Anything change from the last time we spoke? Like what happened on the business side? Because I'm keeping the more exciting stuff about the edge, uh, you know, to, to the next, uh, questions. But tell me more about, you know, the business side.
Dev: Well, well, the business side is exploding. Okay. So we made a mistake, right?
It's, it was a, it was principally a philosophical mistake. We were not really enjoying ourselves. [00:10:00] We didn't fail. So I'll give you an example. We are working with private school groups in the UK now. Per contract is over a hundred thousand a year ar Right. And these contracts go for multi-year because schools don't change software every year.
So we won that and we are continuing to grow US is a very big market, but we are only siphoning that and we have a specific team now who are doing it to 30% of our work because we have to empower. 750 million people. We have to empower those people who we first came to serve right now without going too much into the business angle of it, because there are certain things that we will be launching in, uh, December to talk about it.
Mm-hmm. Let's, let's pinpoint on one factor here, which enables us to do this, and we can also then talk about value. All right. The point is today, if a learner in rural Iraq, right, in rural Afghanistan, in rural, um, code dire whatever, in Africa, right? If they are [00:11:00] using our systems, you know the pain of AI systems, right?
You've got the training costs, that's fine, and you've got inference costs, right? We have near zero inference cost, so. From a business standpoint, we are the cheapest AI specifically in learning in the world today. All right, do we scale? Let's talk about value now. Do we scale and get and serve a couple of million people?
That's what we are doing. Are we supported by people? We are supported now by, uh, you know, UNESCO bodies by un dp, by governments in multiple countries in Africa, right? And we are really going in there to spread rapidly because we do not have a painful cogs. And in certain aspects, cogs by any chance. Okay.
So now let's just talk about it from a value standpoint. We've got an enterprise style business that's working. Mm-hmm. That's growing, and we'll grow it slowly, but with a very strong foundation. But we've got another one where learning can suddenly become, you [00:12:00] know, uh. What can we say? It can be sort of scaled like never before.
Mm-hmm. And this is, this is the holy grail of learning. You're talking about a teacher. It's not just watching videos somewhere. Right. So this is the entire change, and this is because we went back to our philosophy.
Mehmet Gonullu: That's, uh, you know, interesting to hear from you. They have now. What I'm curious about is more the tech technology side of it.
Now you mentioned that now I can run, you know, or I can have the teacher even on a kind of oppo style phone. This is like a very bold statement, I would say. Um, let's deep dive a little bit into this just to hint on something, you know. I think this topic of having, uh, ability to run, uh, not only ai, like any type of, you know, compute heavy application on edge [00:13:00] is something, you know, every enterprise, every company, uh, even from personal use perspective.
People want to achieve this. Now you get this property, uh, AI that you have developed that is able to run on the edge offline. Offline. You're saying this is big. Mm-hmm. What can you share about it? I know like you might not be able to share the nitty gritty of it, but on a high level, if you can explain to us how that works and later maybe I, I, we can follow up on, on the impact of that.
Dev: Yeah. Yeah, absolutely. Okay. Uh, we can share that and maybe at some point, which I'll discuss later with you, I can also share it for you to touch and see. Right. But, but let's, thank you. Yes. But let's just come back to this technology. It's not just the ai, right? This is a heavy engineering piece. Right. So there is only so much you can do with your sort of, uh, [00:14:00] you know, know the AI optimization.
If you're talking about, uh, you know, the, the knowledge element, knowledge distillation, if you're talking about quantization, right? Everybody's trying that. We have peak performance there, but it's just not about that. Our system has 3D avatars. Voice first. All right? Multiple lessons. LM running on a 2020 Android 11 oppo phone.
Right. Wow. That's the breakthrough. That's the breakthrough. And our system is also, uh, sort of multi os. So we are on Android, we are on Chrome for Android, we are on dot e XC for Windows. We are on Lin. Okay, so it's, that's how much we built it and it works right? And we are not gonna actually hide it. We are, we are not gonna do that.
That's why I said the other business models are coming post September to December this year we are gonna launch it as an SDK because we need a part of AI in the world that does not have to bow [00:15:00] down to the frontier models because in a lot of our use cases, you don't need a trillion or a hundred billion parameter model.
It's like driving a Ferrari to your supermarket, right. You need the right use case and AI to do the magic. So we will be launching it as an SDK. We've done a software run already. People love it. Okay. But another, you know, fun part of this is, and that's where we might get stuck, but I really want you to touch it, is we don't have a Mac yet.
Uh, you know, the iOS, the reason for that, of course is because this version of the product is for that part of the world where iOS is not big, but it's got Linux, it's got Windows, it's got Android, it's got Android, Chromebook.
Mehmet Gonullu: De you mentioned something fundamental, which I, I like, it is like. You don't need to over complicate thing as long as you can do the job even on, you know, less parameters.
Um, especially when it comes to LLMs. Absolutely makes sense. Now the question for you, because you know, when we talked in September 20. 23. It [00:16:00] was like only 10 months, I believe, since or like maybe 11 months since Chad GPT came out. And, you know, um, although like you started even before, right? Mm-hmm. Uh, but since then, what happened?
We started to see kind of. Push for open source, you know, LLMs that people can customize and run it on their, you know, own machines. Right. Uh, we saw like a lot of things. So we thought the, the deep sea that, you know, even I remember that week, the stock market little bit and then it reverted back. So my question, how much these breakthroughs that were happening in parallel, justly, like also.
Kind of maybe inspired you to that to say like, yes, this is something doable. We can, if, if Deep Seek is able to do it cheaper. Right. Um, and, you know, regardless of what people said, you know, and their claim, but still they were able to do it on much [00:17:00] cheaper, um, hardware. Uh. What I'm interested, like not the deep seek, as deep seek itself, but you know, this idea of making LLMs open source, you're mentioning about the SDK, how much that played a role in sharpening this vision, you know, that you and I like, by the way, when you said you pivoted a lot, this is a healthy thing to happen, so, but how much that, you know, storm that was happening on the edges?
I would say beside, you know. Sharpened this vision and made you believe that yes, we can do it. We can actually run this full thing, LLM, avatar video sound on any hardware one and or any like, uh, operating system, of course from the ones that you mentioned. So take us through this, you know, this journey of how much that what was happening outside sharpened this.
Dev: To be [00:18:00] honest, uh, it didn't, uh, not a lot. Right? Mammoth? Uh, I, I'll, I'll tell you why. Because we had our first soft launch on Android only before the deeps seek moment. Okay. Okay. Uh, and we were not blowing it too much because we wanted to be able to hit the gas, which we are now. We are. You know, the fastest AI upskilling in Africa, and that's a scalable model for 600 million people.
We'll get to that later, but the point is, uh, it didn't impact that much. Sometimes I see things that are coming out like the new Gemini three N model. We've been working around that for a while. And mind you we're just a team of 20, right? For so on a weight class, we are very good at what we do, but the point also is.
Uh, we don't pivot. At least now with the maturity that we have gotten into ourselves. We don't pivot because of the storm or, or whatever else is happening in the world of ai because AI changes every day. Right. We changed because we understood that this is the pain point. I can do a hundred [00:19:00] billion parameter model.
I can do the best UI and stick A GPT back there, but these guys are not even using Google on a regular basis. The only time they try and get onto the online is when they have to use their mobile wallet to pay. Okay? There's another problem. Some places they say they have internet, but they've got one MVPS doesn't work, right?
So that's the pain point that we are trying to solve. And that's a blue ocean. Why is it a blue ocean? Because if you see everybody's competing, we want to be the next ERP software. We are gonna fight SAP and Oracle. Okay, we wanna do this, we wanna do that. But everybody's above that line. Nobody's in the disruption zone.
This is 750 million to 1.5 billion people who are still not served. And our, our AI teacher before also in the cloud, has served and attained things like 95% CLO course completion rates. So, you know, for the audience, normally an online course completion rate is around seven to 15%, 95%. Now imagine we take this and give it to these people who haven't [00:20:00] even had a proper digital experience.
That's where we want to go with this, and that's why, you know, uh, we were very specific on our needs. Like every decision. Is like, uh, you know, Android first because it's the Oppo phones. Then they have these desktops, not even laptops in schools. So it has to be Windows, windows 10 as well. And then we need to be in Linux.
You know, they're using all of these things. It was strategically planned to do, and luckily we were able to achieve this today.
Mehmet Gonullu: Great. Now let me and play the devil advocate here, uh, with you. Mm-hmm. Um. It's amazing, fascinating to have, you know, everything you have built before to run on the edge, uh, offline and perfectly.
Now, part of. You know, the things that ai, when it's connected to kind of a backend, which is the cloud, is the ability to do two things. First to track the progress, and the [00:21:00] second is to like more personalize the experience. How you would achieve this dev in such
Dev: environment. Very good question. And there's a, uh, answer because we've talked, talked about it.
AI without data analytics is useless. Okay. Mm-hmm. We know that, right? So essentially, again, the data piece, right? Private data not included. So all your private data does not leave your device. We are talking about vulnerable population. The private data never leaves the device, okay? But, uh, that said, we've done our system in such a way that an app also has to update, right?
Mm-hmm. All of these people touch the internet for a couple of minutes to a couple of hours a month. They have to. Specifically in the Kenyan context, right? They have mpsa, of course, they pay online. We can do an entire update and take all of the relevant data for personalization, localization, whatever else in [00:22:00] under three minutes.
So every time they're in the net, in three minutes or under, that swap is done. Right? So we are not sacrificing anything. And that's also an engineering piece because a lot of the, uh, you know, the data collection, uh, you know, the, the, the sort of. Uh, fragmentation, uh, you know, all of those parts which are traditionally raw data is taken up and then sorted out in the cloud is done on the device.
So when we pull, we pull exactly what we need and want, and we push what they need and want.
Mehmet Gonullu: That's just out of curiosity. Have you, um, you know, registered any patents for this dev?
Dev: No, I haven't. Uh, it, it's worth thinking about, but to be honest, in the world of AI and software, somebody will find a different way of doing it.
Mm-hmm. We are just going to, we are, we are gonna do it differently from September. We've got some feedback already. We are gonna push the SDK out. It's, it's better to be at the forefront of it. It's better to support other people who want to do [00:23:00] similar things. Right. We may have some guardrails so that they don't hit us directly in our business, but.
The more people that use it, the better it is. And it's good. It is good to be a leader in this space because nobody can, you know, you, you can raise, uh, $80 billion as open ai and then you're still struggling, right? Meta is coming and taking your top people away. So why do this, you know, patent drama and all of that.
It's better to be in the forefront and keep, you know, if you can get the right minds to support you, you're always at the forefront because it's a collective effort.
Mehmet Gonullu: Cool. Now. This breakthrough, of course, the use case you talked about, you know, the impact touching the, uh, segment, the large segment, 750 to 1.5 billion people, uh, in, you know, different part of the world where, you know, getting the latest and greatest is not possible.
That's fantastic. Now, in your mind, what do you see some other use cases is that can be applied like in addition to, you know. The traditional learning [00:24:00] or teaching, let's call it this way, how I can benefit from this in such economies, which, you know, like, like the African economies or maybe the rural areas of the countries you mentioned.
What else can we get out of that?
Dev: Great question, mama. There are two, two ways I'll answer this right. First, I'll just talk about our model, and secondly, I'll hint on something that you were, uh, possibly hinting on, which I personally find exciting. So the first part is. In this trajectory, we have a model which we've developed.
It's called the C Model, okay? SE, A, mm-hmm. C. The S in the C essentially stands for skills, so teaching, training, upskilling, right? Every country has to teach, train, and upskill their population and also in a direction, all right? That's what our AI teachers do. Now we get data analytics from places. Where there is no data captured for [00:25:00] most of these governments, or it's still in pen and paper, your roll call is in pen and paper.
Your vote is in pen and paper. Everything is pen and paper. We are digitizing that for them. That is data efficiency, which is the efficiency for E, right. And E is accessibility, which is no internet, no problem. We are there. No latest hardware, no problem. We are there. So that's the C model. Okay. So this is not just about the AI teachers, it can help the economy for whole country.
That's why we are working like in, in the Africa context, we have ministries we are working with already, already. So that that's that context. The other side of this is. This, I'll just leave to you and the, uh, imagination of your listeners. But again, just coming back to that point, if they're not using Google regularly or whatever else, right, and they use this regularly, properly, this is their first digital experience.
So if you remember 2013, the next billion lab, I think Google started it or whatever, to get to the next million people. They couldn't succeed because of these bottlenecks. This is [00:26:00] their first proper digital experience now in the West and in UAE where you are. We also know the next wave of digital experience is going to be voice first, right?
What is our system? It's voice first. So can we jump the gun here and that system with some other sort of expansion to it also serve other purposes? Absolutely. But our first target is the educational learning. But if that relationship builds, if that experience holds, we have a lot more ground that we can very easily capture.
Mehmet Gonullu: Absolutely. Now beyond vo, because you mentioned voice first. I tell people I'm, I'm not like a, a fortune teller, but I spot signals and I remember. You know, I said the first, um, time I start to see, you know, the, what open AI 11 labs and other companies start to, to do in that, you know, [00:27:00] space. Mm-hmm. I said like, you know, we, we are going, we are going to the voice first in everything because mm-hmm.
It's the natural thing that we are born with. Like, we, we like to talk, right? Yes. Like we, we, we like to interact using, using voice, but also we like to see. Mm-hmm. And this is the question, and just out of curiosity to have, do you think. You can take this technology Also, of course, I know there's the video part, right?
So the avatar appears on a screen. But what if we want to, and um, excuse me if I'm a little bit here exaggerating things, I want to take it to the next level. I want maybe if I know, I know like the, how you call it, like maybe if someone in parallel works on affordable technology, that allows also the virtual reality.
For these countries. So do, do you envision this, like to have the teacher, uh, that you know, the artificial intelligence teacher that you have? So me, a child in [00:28:00] Africa, or a child in maybe Cambodia or, I don't know, in a rural area. So I put that glass and I see my teacher. Am I, am I too much futuristic here?
Dev: No. You know why I'm smiling Mammoth, and if there is something more to discuss on this, I would really love to discuss with you. Sure, please. One of the things we've learned also is hyper-focused. That's what we are doing, but in November of 2024, we have already built, we are with embedded LLM. We have done it.
So what you're saying, we already have it, but is that the, uh, we have hardware constraints in these countries, right? You, if you remember the bottlenecks, the bottleneck ly, yes. Connectivity in hardware. So that's parked at the moment. The moment we can get that in. It's ready.
Mehmet Gonullu: Yeah. This is why I ask you Dev, because I'm sure this discussion, and I'm hopeful.
I'm hopeful that. You know, you know what? Like people comes and [00:29:00] tell me sometime, you know, like, what else can we do? Like, uh, everything is like, uh, already there still. We have a lot of problems to be solved and I always refer them to. I would remember the name of the book very, very, very shortly. And there's another like thought leader who keeps saying, you just keep need to find these problems.
And if you can't solve 1 million people problem and just sell it for $1, here you go. You have a $1 million business. And for me, you know what you just mentioned, for example, look at what you have done. The still we have the. Challenges of getting proper internet, of getting, um, you know, hardware affordably.
So I think that's maybe also another space for people to innovate in that you did it on the software layer, which is fantastic, right? So, uh, because now it's not possible. Maybe we still need, I don't know, I'm just making up the number. Maybe we'll still need five years, but we don't want, so what you have done, I don't want to wait.
Five years [00:30:00] until computing becomes affordable in Africa. You want to say something I think Deb, right?
Dev: Yes. As I said, uh, we are hyper-focusing, but if there's an opportunity, it's a podcast or whatever, I'm just saying let's have a, have a conversation around it. 'cause I'll just give you a little bit around it.
Right. Africa, that's where we are going. Right? And we are going to schools and all of these places now. That's great and this will come there, but this is what this technology does. The one we have developed for ar vr. With embedded ai, it's just not about you having a teacher. Right. Which makes sense because AI is about hyper personalization, right?
The holy grail of learning VR and AI is about immersiveness the holy grail of learning. You put them together and that's the perfect piece. Okay, that's fine. But I'll just say the different thing. Because it's a technology show, what we can now do is if you just think about a traditional video game, right?
Mm-hmm. In a traditional video game, you have a level. So you have [00:31:00] characters, you have to choose one of three dialogues and you take your level path, right? Right. We've built child's games in MVP format, which I. You have endless possibility of conversation. So you can have a game today with a character with an LLM in each character.
He can do that. Right? Wow. So, so it, it, it, it's done. It, it, it, and it's in the same trajectory, right? It's part of that layer because we control. Right from fine tuning of course, because of all of that. But our edge layer, our rag layer, our, you know, uh, uh, the, the optimization piece, which I can't say too much about that layer, but we are essentially six layers deep in terms of our, uh, research, which these two products,
Mehmet Gonullu: and I think dev, there is no limitation in, you know, what kind of topics I can go and pick to learn.
Right?
Dev: No, no, no. You can upload
Mehmet Gonullu: A PDF and it'll teach you that. No, because my idea is, you know, if, if [00:32:00] also we can put in the hands of, um, the population, the mass, and teach them also about, let's say business entrepreneurship, it's not like about the traditional education in the sense of, uh, what you, what they call it, the stem.
Don't get me wrong, STEM is very important. But to teach, you know, people about these, um, other skills that, in my opinion, and I think maybe we touched on it in the previous episode we did, is, you know, teach them how, how, how to run a business like, right. So, so, uh, today we started to see, you know, this, um, solopreneurs and vibe coders and, you know, all this.
I would love, you know, to see, you know, people from. Every continent, every place on earth that are able to know and learn. Of course, they need to learn. They need someone to teach them. And this is where you come into the picture Dev with what you have built. Because maybe, look, you and me, we were lucky and maybe we had mentors in business that came and teach us, but [00:33:00] imagine if you put this in the hands of, of the mass population.
Like wouldn't be nice you think.
Dev: Yes, absolutely. I fully agree with you. I'll just give you two examples. Sure. Which may, yeah. Which may show you that we are already doing this in, in a certain degree. Okay. First of all, let me give you an example. In schools, we obviously teach the UK curricula. We teach the Kenyan curricular, we teach whatever curricular, right?
So first it's localized. It's not a GPT. You don't know where the information is coming from. Your curriculum, university, your lecture, your curriculum, right? So that's fine. That's the first part. The second part is the breadth of this. We've taught law, psychology, et cetera, hard skills, right? But our single biggest at the moment, uh, hit if you wanna call it, is employability skills training 17 countries and counting.
You see how important that is? Are we doing entrepreneurship? Yes. Are we doing uh, sort of green, uh, education in that own way for farmers, et cetera, to understand global warming and all in Kenya? That's what we are looking at with vet ministry. Now we are doing AI literacy on the [00:34:00] UNESCO framework, so it is content agnostic.
Okay, so that's essentially that part, uh, covered. Now the second part also I'm saying is of course this system teaches you. Alright, so let's take a business example. This system teaches you, you've learned business, you've started a startup. Mammoth is your investor. Let's say that, okay, now you've had a board meeting and mammoth's not very happy.
You can talk to the same AI teacher and say, Hey, this is what happened. You know, this is what we've learned. This is the data. What else can I tell Mammoth? It'll tell you it's the same thing, right? You just, it's not only teaching you, it's there with you 24 7. The same thing for your, no, just like before an exam, you can ask, Hey, tell me about this, or this scenario is coming, what could be the answer?
It's the same for entrepreneurship. It's the same for whatever else. So it starts as a teacher, but it stays with you, right? That that's that extension piece
Mehmet Gonullu: because you have the engine built already, so you can load whatever it is, right? It
Dev: knows what you have loaded before and it [00:35:00] knows you because it's talking to you every day.
Mehmet Gonullu: Absolutely. Yeah, yeah, yeah. This is, this is really fascinating, dev. Now, you mentioned about some, you know, upcoming, um, upcoming like things in, in the few months from now, what's now your, I would call it your, your long-term vision dev, where you want to take this.
Dev: See, we, we've always built this for the long term, right?
Uh, if anybody ever asked us for a quick sale, we would say no. Right? Uh, we still hold that. That's why we've struggled so hard and work so hard. Where are we going? We are hyper focusing. So we still work with some universities, but that's not where we are hyper focusing. Uh, also because in the west where they can pay, they don't have their policies in order yet, they still don't know what to do with ai, right?
So where are we focusing? Our enterprise model is absolutely focused on private school groups. So UAE we haven't explored. We know there's a great opportunity there. UK is blasting special education in private skill, [00:36:00] uh, school groups. We are the best in the world right now because we've got that much deep.
We've built the same part of our product with the UK's second biggest, uh, sense special education, you know, private school group for nine. Right. So that's the, that piece. In terms of our larger 70% piece, which is, you know, close to our heart, we are really focusing on Africa right now. So in August, we are going to be upskilling 25,000 mostly marginalized learners in Kenya on AI literacy.
Right? Mm-hmm. We are gonna stick to that because we already have two county governments who have come in and we are being challenged to do 50,000. Let's see if we can do it. That's already there on. At the same time, we have our government partnership with the government of South Tome, which is on the other side, Africa, smallest economy.
Right? It's on the west. Same time now come September. It is bound to be a success. We need to obviously make it better or whatever, right? Come September, we will now be scaling this. To the 47 counties of, uh, [00:37:00] Kenya, right to Guinea, besa and other countries in the west, and then get into the mainland properly.
So it's a east, west attack. Now, the best advantage we have is if word of mouth kicks in because it might, right? If word of mouth kicks in on top of it, we have the UN bodies, we have the government supporting us. This could really explode and explode in such a way where we don't have inference costs. We won't hit that bottleneck that, oh, uh, what's our sort of cogs?
What's our cogs? We can handle that. So that, that, that's really the exciting part here.
Mehmet Gonullu: Absolutely. Uh. So this is for the audience. Like, as you can see, it's the AI meeting education, um, or ed tech, if you want to call it, at its best like that, really. Um, I, I like, you know, the energy. It was there before, but I can see, you know, the breakthrough that you've done.
Um, it's fascinating, you know, coming from a, um, uh, like engineering background and having the ability [00:38:00] of running something without, because, you know, I grew up, you know, and the internet was not something affordable for me, you know, when it first came. And I understand the struggle and I know in economies, especially like Africa, there's plenty of talents and all what they need.
They need the tool, right. Okay. And if you don't have, and I like what you said, you know, like if it's not there, we come into the picture. Right. So, so you're offering that, that's really fascinating. Uh, anything that you want to highlight, anything maybe I missed to ask you?
Dev: No, I, I, I think it's an exciting time.
Uh, I don't know, you're very busy Mammoth, but maybe we can speak September, October once, because we'll have the results then with
Mehmet Gonullu: pleasure. With pleasure, yes.
Dev: And uh, I, I think the other thing is just piggybacking on what you said, we will be there. Let me just explain that a little bit differently again.
Sure. Everybody, everybody's fighting. Above the line. Okay, good for them. I don't wanna fight there. Uh, so when you're talking about below the line, this is this opportunity, right? I'll [00:39:00] give you the example of Vietnam. Vietnam never really had line line phones. They jumped straight to the mobile Right.
Because of the war or whatever. But they are at par, right? Right now they have better internet suites than some parts of the uk. This is that opportunity because you need that teacher in the equation. Just giving books, donating books, and whatever else won't work. And this is that opportunity. So if this AI literacy scales, we are not gonna stop them after that.
They're going to get AI literate, but they're gonna get other lessons and other lessons and other lessons. So yeah,
Mehmet Gonullu: absolutely. Where people can get in touch deaf, if someone is listening today, whether they want to be part of this, whether they are interested to test that. Or they want maybe to, to help opening doors also as well in Africa or elsewhere.
Dev: Thank you. Thank you, uh, for that opportunity mammoth. So you can just write to me directly, uh, it's dev DV at the rate oi ED u.co uk. So Dave at the rate OI edu.co uk. [00:40:00]
Mehmet Gonullu: Great. Thank you so much, dev. Really, it's exciting times. Uh, you know, a lot have changed since we spoke in September, 2023. Uh, and as you said, you know.
The AI changes every day. When, when you mentioned Yeah, absolutely. It changed very fast. The, the people talks about, you know, negatively, generally, and I tell them, guys, let's just ignore everything and look at the bright side. And I think what you're doing, dev with the team, is something fascinating. It shows the power of technology, the power of AI being used in goodwill, being used for good causes.
For impact. I love it. Just, you know, I, I don't, I don't do that much on the show, but I, again, like I, it's something close to my heart because impact, um, enabling, uh, developing economies, getting the next generation of. People who are literate in AI and other technologies business, as you [00:41:00] mentioned. It's really fascinating.
Thank you very much, Deb, for being here today, and I will make sure that I'll put the links in, in, in the show notes or if they are watching this on YouTube. Thank you again for being here, RU Today. Dev. As you know, before I close, I have to speak to my audience, so this is for the audience. Guys, if you just discovered this podcast, thank you for passing by.
I hope you enjoyed it. If you did, so please give me a favor and subscribe. Share with your friends and colleagues. As you can see, we are trying to do an impact with guests like Dev, who are building fascinating technologies for the next generations, and if you are one of the people who keep coming, again, again, thank you very much for your loyalty.
Thank you for your support. I'm really, really grateful. I know like I'm repeating this in every episode, but recently the podcast is doing fantastically well. We are being on the top 200 Apple Podcast charts in multiple countries. The latest I checked it was eight. I never seen this in the journey since two and a half years, it can [00:42:00] happen without two factors.
First, you, the audience, thank you very much, and second, my guest, including you, dev. And you are a second time guest and I'm sure we have, we're gonna have your third time also as well. So thank you very, very much, and as I say, always stay tuned for a new episode very soon. Thank you. Bye-bye.