The conversation this week is with Panos and George Karagiannis, co-founders of Moveo.AI. Both of these guys are passionate about new technologies and always eager to learn. They love turning ideas into practice by working with people that share the same enthusiasm about technology as they do at their company. They are building a new type of conversational AI platform that works on automating mundane tasks, using no-code platforms and self-learning capabilities to transform customer experiences.
If you are interested in learning about how AI is being applied across multiple industries, be sure to join us at a future AppliedAI Monthly meetup and help support us so we can make future Emerging Technologies North non-profit events!
Resources and Topics Mentioned in this Episode
George Karagiannis 0:00
Open AI which of the three is going to make it very easy for new engineering teams to get started, you know, things that people will take like three, four or five years to even like create. Now you can do it very easily. One concern that I would have with outsourcing your AI would be data privacy, especially GDPR European rules, things are going to have to get sorted out when it comes to GDPR. Because things are kind of strict the Europe is going to it's going to be hard for engineering teams to ensure data privacy over European companies.
AI Announcer 0:31
Welcome to the conversations on applied AI podcast where Justin grams and the team at emerging technologies North talk with experts in the fields of artificial intelligence and deep learning. In each episode, we cut through the hype and dive into how these technologies are being applied to real world problems today. We hope that you find this episode educational and applicable to your industry and connect with us to learn more about our organization at applied ai.mn. Enjoy.
Justin Grammens 1:01
Welcome everyone to the conversations on applied AI Podcast. Today we're speaking with Panos and George co-founders of mobile.ai. Both of these guys are passionate about new technologies and always eager to learn. They love turning ideas into practice by working with people that share the same enthusiasm about technology as they do at their company. They are building a new type of conversational AI platform that works on automating mundane tasks, using no code platforms and self learning capabilities to transform customer experiences. Thank you, panelists and jurors for being on the show today. Hey, Jason,
George Karagiannis 1:32
thank you very much for having us.
Justin Grammens 1:34
No, I'm super, super excited, because you guys are located halfway around the world pretty much from where I am here. So thank you for taking time to spend with us from Greece. You know, I gave a brief intro about your company, maybe Georgia, if you want to go first, maybe talk a little bit about your background, and maybe the trajectory of your career that got you to where you are today,
George Karagiannis 1:51
of course. So by nature, I'm a tech guy. So I'm an engineer. For the past like few years, or have been dealing with mobile, I've come to learn, you know, new, like skills around like administrative marketing, things like that, like running a company. But my background, I did like real science in UC Santa Cruz in California. And then I was a PhD candidate at Cornell, I ended up like living with a master's degree. And I have a few like, you know, my years of experience in the industry before starting up mobile. My background is mainly on NLP and databases. So this is one of the main reasons that we started Moveo, because you know, we had a pretty good and pretty solid background are the technologies that we offer to our customers.
Justin Grammens 2:38
That's cool. That's cool. So So you studied in the US. But now, I decided to start your company over in Greece, right? So
George Karagiannis 2:44
mobile is kind of a has a kind of a hybrid theme. We're kind of split between New York and Athens. We're also expanding like internationally. So we're making a few, you know, first few steps in Sao Paulo and Brazil in the Brazilian market. So you could say that we like split between the three parts of the world. So often Sao Paulo and New York.
Justin Grammens 3:07
Oh, yeah, as we talk a little bit more about the product and the technology and stuff be curious to jump into, like how it's been to sort of manage those teams, but I don't know. So you'd like to give a little bit here.
Panos Karagiannis 3:16
I'm also you know, coming from an engineering background, I also studied in in the US, UCLA, I got my bachelor's and then moved on to UC Santa Cruz to do my masters. And then I worked for a couple years in Boston, in the industry, where I was mostly doing, I was of the on the intersection between research and product. So basically, was working on the scalability of let's say, NLP algorithms and how these, you know, new technologies, even though they're very efficient in the lab, and they, you know, give out such amazing results in research, how can those algorithms be used in real life, and that's all before starting mauviel, where, you know, setting in your company, you have to wear many hats and juggle many trades. So we have to learn a lot of things as we go along, you know,
Justin Grammens 4:01
yeah, for sure. I mean, the life of a startup and obviously, the life of any new technology, you guys just sort of pushing the envelope on on what's possible in the forefront. It feels like you guys, were maybe I guess, correct me if I'm wrong, and either of you can answer this question. But, you know, you were you were brought up in academia, you know, you learned a bunch of new techniques, especially around NLP, which is really fascinating area, but yet you maybe it wasn't scratching that itch with regards to like, I actually want to build something that is going to change the world.
Panos Karagiannis 4:26
Exactly. It doesn't that's actually very accurate. You know, and but I think that applies for both me and George, we were considering following a path, a career in academia. But then we realized that, you know, in academia, even though you deal with some very smart people, there's also this culture of publish or perish. And a lot of those research papers and the things that you do, even though you know, they improve the algorithm by a tiny small factor is not really applicable to real life. And I think that's what kind of interests both me and Joe It's how we make things applicable in the real world. And that's why we moved from academia to building things that scale. And they, you know, they make people's lives better.
Justin Grammens 5:11
That's excellent. Yeah. And you know, after using the tool, I'm really impressed with ease of use, it's free to use, right? So I downloaded sort of started playing around with the free trial. And it feels like you guys were really sort of focused on sort of giving the tool to the end user to do however that they want. It's not like you, as I mentioned, at the beginning, it's a no code environment, right?
George Karagiannis 5:31
That's also one of our big, like, challenges that you have to face from day one is how do you take like, complex algorithms and complex NLP models? And you make them you make them available to like end users who have no idea, not only about computer science, but like an NLP, but like, you know, about, like basic tech stuff, you know, our users, right? Some of them are tech savvy, but some of them, you know, they come across with concepts for the first time, the challenge that we had from day one, is how do we make it possible for business users who have, you know, the capacity to learn and know the processes of a company to actually put them into practice, and parameterize, a system that uses NLP in the background, that is going to be able to learn from these inputs, and be accessible and available to the end users of a company. And it's going to be able to automate and solve real problems. And you know, this is a challenge that is ongoing, like every day that we wake up, we think about how we can make it easy for the person who's building the virtual assistant, or who's building the AI automation, to do so with the least friction possible. And this is a big, big challenge.
Justin Grammens 6:57
Yeah, for sure it will. So I kind of jumped right into the meat of the product, maybe you guys could back up and maybe tell me a little bit about what the product does. And maybe some specific use cases you're working on today with with customers.
George Karagiannis 7:07
Basically, what Moveo is, is a SaaS platform, right? Which is accessible to everybody to sign up and use just like you did, Justin, we're b2b. So we targeted companies that are in need of automation, and are in need of automation, when it comes to solving like inbound requests. So you can imagine that, you know, you have questions or you have multiple tickets, not only FAQ's, like, you know, what is your pricing or what is your operating hours, but you have queries that, you know, require automation that needs to actually take action, for example, you may, you may want to schedule appointments more efficiently, or you may want to, like reschedule your packages you like any show, and you may want to like reschedule your packages efficiently and automate this process. So we're pretty much in the automation business to be very broad. So this means like scheduling appointments, rescheduling packages, answering, like, frequently asked questions, and many, many different use cases. And we provide a platform, such that like, from which, like customer companies can log in, sign up, import the data, in a no code UI, describe their business, just like they would describe it to a new hire, and what they get out of it, or deep learning NLP models that can actually, you know, get the job done and automate, like, from 60 to 80% of the incoming requests that they get from their end users.
Justin Grammens 8:35
That's awesome. Yeah, and, of course, with the shortage of, of workers these days, and yeah, you know, as I talk to a lot of people, you know, where can AI benefit, it feels like it can benefit companies, as it sort of, I say, it's not really like going to replace a job per se, it can actually take take the jobs that people don't want to do number one, and then it can actually augment them, right. So I'm guessing maybe some of these companies are using this for initial contact, but yet, if something needs to be bubbled up to the to like tier two, or something where human would take over, then that's where a human would be useful, but some of the initial automation can be, they could use your tool for that.
Panos Karagiannis 9:10
Exactly, Justin. And well, we've seen that a lot of the companies, you know, except for the financial benefit with this, like all views, what they also achieve with a tool like that is that the mundane tasks, they don't have to be, you know, handled by a customer support agent and that that a workforce can work on, you know, developing other skills that, you know, that are much more valuable for the business. Plus, you know, you get all the other perks of having a digital employee that's 24/7 available, can handle peaks during Black Friday, I assume Amazon probably had a huge surge in in requests and then delivering packages. So you know, that scales like that as much as you want and then, you know, you have a bunch of benefits when it comes to customer support.
Justin Grammens 9:54
Yeah, very, very cool. There's a lot of companies that are doing this already today. You know, how are you guys differentiating yourself? Do you feel against some of the other competitors out there? Because you guys are just sort of getting going here? And?
Panos Karagiannis 10:05
That's a very good question. So just in, in a natural, we have tried to kill the learning curve, in order to use our tool, you don't have to learn anything new. So all we ask you to do is really know how to speak and how to write. So you can just log in, and you can describe your prop your problem in English. And then you click one button. And right out of the box, you get a virtual assistant, that basically, you know, it's gets adapted to what you just described. So you don't have to learn, let's say, abstractions that are, you know, related to building a virtual assistant, like how the dialogue works, or how you train the model, if we have tried to make it, what we call in a broader sense, auto AI, you just know how to speak, you know how to express your problem, you type it down, and then we generate something for you, that's very, very intuitive that you can deploy with the click of a single button on your website. And on top of that, let's say you have historical data, let's say you have transcripts from past conversations, with the click of one button, you should be able to upload those data. And you know, in a few hours after the system trains, we have the virtual assistant ready for you that you can deploy again, on your multi channel, right? It can be WhatsApp, it can be on your website, Facebook, and so forth.
Justin Grammens 11:32
Yeah, that's what I was gonna ask. So the whole deployment side of it, I use your tool and I generate these models do, do you guys then support an API that I call into? Or do I grab a widget and put it on my site that has, you know, your Sass product? Like, what are the next steps, once I've kind of build something,
George Karagiannis 11:49
once you actually build something, we'll make it as easy as possible to move forward, which means that you can either like grab, like, a few like, like lines of code, like a script that you can put in your website, or you can connect your favorite like communication channels, like your Instagram, or your Facebook Messenger, or, you know, your whatsapp, things like that, with a click of one button. So we try to make it easy when it comes to how you move forward. One of the big pain points that companies have to deploy a virtual assistant in production is how do you actually improve it? What are you know, like, why do you actually build this thing that, you know, why do you know that your customer is askable, you don't really know, you know, you have some bias, you have a prior idea of what you know your customers ask about. But when you actually deploy it, and people use it, they may actually ask different things, or, you know, what you thought about is just a subset of the things that they ask about. So one of the things that we have worked very hard at maveo is, we call like a technology called self learning, which takes like the chat logs, the actual conversations from the end customers with a virtual assistant, and manages to learn new stuff that people ask you, and incorporate this knowledge into the virtual assistant with as little friction as possible. And we have done this, and we have done this successfully. And we have many success stories that you can share with our customers that have actually used this technology. And what they have actually accomplished is that they have avoided looking at like 1000s of chat logs with end customers to see where the dialogue breaks, or where people ask different things. Because the algorithm can do that now. And to try to limit the amount of hours that a human has to look into the data and actually make take decisions. We want to limit that. Because human time is, you know, money.
Panos Karagiannis 13:44
It's precious. Exactly.
Justin Grammens 13:45
That's awesome. No, that's great. That's great. Very, very useful, I guess, for business. And you guys are sort of giving this free trial on ramp where where's sort of the breaking point, like, where do we get to a point where actually or what features do I need? How are you guys figuring out sort of like the next step, then to monetize this?
George Karagiannis 14:01
Right? We're still working on that, right now we'll have like a 14 day trial, that you can do whatever you want. And later on, you have to pick your your plan, depending on your needs, basically, you know, to just cut a long story short, depending on the volume that you expect, you pick a different plan. So if you're like a big corporation, and you expect like many people to talk to you, you probably should go for enterprise. If you're like a smaller company like an SMB, you expect like, you know, four or 500 like users per month to talk to the virtual assistant. usually go for one of the you know, starter or like pro plans that we offer.
Justin Grammens 14:39
Makes sense. Makes sense for sure. You know, you guys mentioned it's sorry, YouTube are the founders you have a team of people that are working with you, I guess are you guys growing like what what's the sort of the day in the life I guess if your business where do you see yourself going the next couple of years.
Panos Karagiannis 14:52
We started a couple years ago, Dustin he was he was a very few people, initially me and George and then we kept growing and growing. So right now we're a workforce, a team of 25 people, most of most of the people are based in Greece. But then we have people in Sao Paulo, Brazil and New York in the US. So right now, you can imagine that we have the team that's working on the development of new things. We have a team that's devoted into marketing, we've spent a lot of energy on the UX and the UI side of things. Because as George mentioned, I believe that companies that deal with artificial intelligence in general, they need to innovate on two vectors on two fronts. One is the technology itself, and then is how you serve that technology and that complex ideas to the end users. And that has never been done before. All right. So we constantly need to innovate. So we try new things. We have engineers that are devoted to r&d, using the latest and greatest things like CPT, three, and so forth. The country is very close to normal startup, we have our scrum meetings. It's a very collaborative environment. We have our ping pong tournaments, our pizza Friday. So we have fun stuff as well. But everybody here is, you know, passionate about what they're doing. The collaboration is vital lifing in those first few years of the of every company is great.
Justin Grammens 16:19
Are you guys self funded? How long have you been? I mean, you've been going for two years now.
George Karagiannis 16:23
Right? The strange thing about us is that we have been cashflow positive from day one. And you know, not many startups can say that. And that's why I'm saying it's kind of strange. The thing is that we recently recently raised the pre series, which you know, had a goal of like, you know, helping us expand to new markets, because right now our customers are mainly in Europe. And we want to expand to the US and to the Brazilian market. So this is why we raised the pre seed round. And slowly and steadily the more we improve on the product and on our technology offering, we might be considering of raising series A or A seed round, soon ish, I would say we don't have a specific time plan,
Panos Karagiannis 17:07
and no pressure also because we're not burning a lot of money. So we are cashflow positive. So we're calculating our next moves. Things are growing fast. Now right now we're serving millions of users across 85 countries. So from we get requests from different geographies, so we were just calculating our next moves. And we're very hopeful about the future.
Justin Grammens 17:30
That's great. Are you guys you guys hiring these days, I'll be sure to put a link off your careers page, anything in particular you wanted to maybe stress that you're looking for, I was gonna
Panos Karagiannis 17:39
say as long as you're passionate about technology, you can you can apply to mobile, we're very open to new people, new ideas, and I think it's gonna be great, a great fit for us, as long as you really like what you're doing. And it's going to be great.
Justin Grammens 17:53
That's good. That's good. I liked that fact, I think there's a lot of need for people that are more creative and artistic, maybe already, you know, maybe aren't deep into the technology that actually because like you said, it feels like it's you guys are your mission is to make it very easy for somebody just to jump in and do this stuff. So you got to have that mindset, that skill set, rather than, you know, being a machine learning expert. You mentioned GPT. Three, are you guys? Are you guys sort of writing on top of some of some of those API's and stuff? Right?
George Karagiannis 18:19
Yeah, that's a big one. So I mean, everybody's going nuts and crazy with Jupiter three. And, you know, it's not new. I was kind of surprised that, you know, people are discovering it right now. It's been around for a year now. It's amazing. What it can do is truly amazing. And I think that there's going to be many revolutions across different parts of how AI is applied into our real lives. chatbots are going to be one of those, like verticals, I would say. And as Pano said before, the interface in which humans interact with GPT three is going to be of vital importance, because people have not paid attention to that anymore. And open AI, which mp3 is going to make it very easy for new engineering teams to get started, you know, things that, you know, people will take like, it will take them like three, four or five years to even like create. Now you can do it like very easily. One concern that I would have with like, just outsourcing your AI, with open AI maybe will be their privacy, especially GDPR and European rules, things are going to have to get sorted out when it comes to GDPR. Because things are kind of strict in Europe, it's going to it's going to be hard for engineering teams to ensure data privacy of European companies. Other than that, I think that movie is is gonna maybe use for some benchmarks to be three and try to integrate Gibidi like technologies into our products for sure.
Justin Grammens 19:41
That's a great point. Yeah, I mean, I built a number of different systems and leveraging third party out you know for what they do best, especially when you're getting started is a great way to go right there's no reason to sort of reinvent the wheel on this stuff but so but but know what you're getting yourself into, especially around data privacy, who is your main point
Panos Karagiannis 19:59
that and data privacy was also add, you know, as I'm thinking now, scalability could be one more, because I know there's, you know, rate limitations and things like that. Those are very, very heavy models, right? The Vinci's, for example, has billions of parameters, which is hard to scale. So for example, you can imagine that during peak time, Smallville gets close to 50 requests per second, when you know, people are really conversing and asking things. And, you know, handling that level of parallelism is a very hard engineering challenge. That, you know, open AI, for example, could easily become a bottleneck, for instance, but things improve all the time. But that's also something that you need to take into account when you're building things at scale.
Justin Grammens 20:43
Good point. Thank you, Panos and George, for being on the podcast. Is there anything else that you guys wanted to talk about? Maybe that I didn't didn't cover during our discussion?
George Karagiannis 20:51
We both think that, you know, we live in super exciting times, and especially for NLP and you know, the timing of the podcast is, is very, very good. Because, you know, ChatGPT was released, like, you know, a few days ago. So everything's press, we think that more technologies like GPT are going to come out, it's seriously going to change the way we live. This is not an exaggeration. chatbots, you know, like, they came out like, nine years ago, 10 years ago, they over promised and under delivered. But now, not only with GPT, but like technologies that are you know, us for the past two, three years now, are seriously going to change the way we talk with customer support agents, and how end users are going to be conversing with companies and getting you know, their their questions answered, then their tasks done. So we're super happy that we have started Moveo, where we truly believe that it's a it's an amazing opportunity for everybody to be spending time with NLP. And specifically for people who want to start, you know, their careers in NLP. It's super exciting things.
Justin Grammens 21:57
Yeah, as you sort of talked about GPT. Three, I was wondering, also, you know, some of the text generation side of it, you know, you know, as as well, so not only the understanding of what's going on, but also just being able to have it automatically write stuff. Do you guys, I mean, maybe I mentioned a little bit earlier, maybe this last question, but like, you know, where do you guys see your technology, then going in the next three to five years, you say it's going to change, you know, and improve, but it's, is it almost in some cases where the humans aren't even gonna be needed anymore,
Panos Karagiannis 22:22
In a sense Justin maybe, you know, but you know, there's many challenges that still need to be solved with GPT, for example, hallucinate or it can tell you things that are not true. 100%. And, you know, though, exhibit is amazing, it's pretty, it can be used on many use cases when you're playing with it. But can it be used in real production, for example, that is going to be the challenge that's going to be sold in the next couple of years. And that's going to completely revolutionize how NLP and virtual assistants work and change our, you know, day to day, basically, on customer support and on many other fronts. But if you ask where Marvel is headed, I would say that what we are trying to do is basically to abstract away all this difficulty right now that someone needs to know a few core concepts about NLP, like class imbalances, how training works, you know, technical terms that normal people don't have to know. And we don't want them to learn. So it's going to happen naturally is going to happen by clicking one button and is going to work. So that's going to be the amazing end product.
George Karagiannis 23:32
Yeah, there's many modes, GPT can be used. I think, for now we're gonna see a lot of the copilots kind of like use case where it's right next to a human, like a GPT is right next to a human telling him or her what to do. So in a co pilot kind of mode, but once he really gets like, 100 times bigger, we're gonna have to seriously talk about like, what what stuff are going to be done by humans and what's up is going to be done by GPT three.
Justin Grammens 24:01
Oh, great. I guess that's gonna be a good reason to have you guys back on the show then for sure. So, bounce and George, I thank you guys so much for being on the program today. I appreciate you sharing everything that's going on at moveo and I wish you guys the best in the future. Look forward to keeping in touch
Panos Karagiannis 24:15
with you guys just and it was a pleasure talking to you.
AI Announcer 24:19
You've listened to another episode of the conversations on applied AI podcast. We hope you are eager to learn more about applying artificial intelligence and deep learning within your organization. You can visit us at applied ai.mn To keep up to date on our events and connect with our amazing community. Please don't hesitate to reach out to Justin at applied ai.mn If you are interested in participating in a future episode. Thank you for listening