Serbian tech entrepreneur Djordje Dimitrijevic envisions a future where technology enhances daily life. In 2023, he founded SignAvatar, a startup dedicated to developing AI-based software that automatically translates speech into sign language. 

Recently, the young startup demonstrated its innovative solution at major transportation hubs across the Balkans, including the Nikola Tesla Airport and the Belgrade Central Railway Station.

In an interview with IT Logs, Dimitrijevic discusses the transformative impact SignAvatar aims to achieve, the software’s immense potential, and how its testing phase in the Balkans is poised to pave the way for expansion into larger markets like Western Europe and the US.

IT Logs: Can you briefly explain the concept behind Sign Avatar and how did come up with the idea?

Djordje Dimitrijevic: SignAvatar is an AI solution that translates speech into sign language automatically. We’re using it to make the world’s first fully accessible PA (public address) system. We’re translating all the audible PA announcements at airports directly into sign language and two written languages in texts, and we display them on dedicated screens around gates, and also stream with their browser as well, so you get access from your phone.

It was inspired by my brother who is hard of hearing and although he doesn’t speak sign language, we had a deaf friend. All of our co-founders are AI enthusiasts, so we patched things together. At first, it was curiosity, it was just a side project, then later on, it became a real thing. We got pushed by some big Serbian entrepreneurs to venture into something new. It was January 2023, a year and a half ago, that we started working on the solution actively. 

We were awarded three different grants, one from Startech (Philip Morris), the second one is UNDP, and the third one is the Innovation Fund of Serbia for GovTech. And also we have a client, which is Serbia Railways, and we hope to onboard Nikola Tesla Airport soon. TransportSign is currently at a pilot stage there and hopefully, by the end of September, we’ll have Vinci Airports as our first client.

Tell us a bit more about how the project with Serbian railways started, and what have you seen so far in terms of results?

In October 2023, we were the premier innovation that was showcased by the government when they opened the Belgrade Central Station. Our innovation was the premier one, and we got an opportunity to pitch it to the board of directors back in May. And then we patched things up by October to open up the station and have everything working there. Of course, it did not work perfectly, there are always some mistakes, and that’s why we piloted there. 

Like all startups, we iterated, we talked to the deaf community, we had a lot of different scaling issues, and fixed all of those challenges. The software itself changed a lot for us, and also, since working with the railways, we got a chance to employ two deaf people to work with us full-time. 

What is the progress of the pilot project with the Belgrade airport right now? 

We also had an opportunity to showcase our solution at Belgrade airport. Right now, we have our software, our current solution, or pilot, on five different screens. During commercial implementation, we hope to have around 30 to 50 screens all around the airport with the fully working software. It’s kind of a demo so we’re showcasing to the people what it looks like, how it feels like, and similar things. We also had a conversation with TAV Macedonia, and we hope to have them as a client in the future as well since they seem really interested in what we’re doing.

What is the main concern when we’re talking about this sort of infrastructure in the whole region, and accessibility? 

Accessibility is perceived very cautiously in the region, as the market for accessibility is not that good. We have a unique business model that enables us to be a revenue generator for our clients, instead of an expense, which allows us to penetrate faster than a normal accessibility company would. 

What we concluded from a bunch of conversations just selling accessibility is not enough. The CSR/ESG budgets are enormously small, or the sales cycle is enormously long. So there’s no real motive for accessibility in the Balkans and if you want to sell it, you have to find a way to bring another way of value into it, besides accessibility. Selling accessibility in this region is one of the hardest things you’re going to sell because in the eyes of the beholder, it’s usually seen as an expense and our company shares completely different values.

Therefore, our expansion plans are not in the region. We want to onboard a couple of clients here and start iterating the product, to make it perfect and then venture out into our primary markets such as the US. The secondary market is Scandinavia, especially Finland. Then third it’s Dubai and similar MENA regions, countries which have a thing for innovation, not just accessibility.

That’s our honest viewpoint of the market right now, and this is only our first product, since we also plan to venture into other verticals as well, especially the second one being education and similar branches.

In terms of developing such a product in Serbia, is there enough support, enough IT talent etc? 

I know four startups in Serbia that are doing accessibility technologies, so there’s something here – people like helping people, and my motto is using the knowledge that I have in technology to help somebody live their life better. And in my case, it was accessibility. So in my surroundings, if someone has a problem – I know how to do tech, I know how to do business and I am the one that will be able to implement that technology to help somebody else. And that’s what satisfies me personally the most.

Regarding the talent, every single country in this region has amazing high schools/universities, especially the top 5% – you get a lot of people who know their stuff, especially IT-focused; AI machine learning, or just general software development frontend and backend,

Coming from me, I’m surrounded by people who are really good at their jobs. So for me, getting more talent, getting access to the talent pool, is not that hard, especially because one of my co-founders is a researcher and a professor at a Belgrade University, so we are also sourcing talent from there as well. The crew that started the company is engineers, so we kind of pivoted a little bit more into business along the way, but we still know how to do our technology.

What do you think is next for accessibility in terms of AI and other emerging tech?

I actually saw a bunch of good AI accessibility solutions along the way, but honestly, the best app that I’ve seen for accessibility uses no AI. It’s called Be My Eyes.

We are an AI solution, but we don’t do stable diffusion (deep learning model) and we actually have a good reason to think that stable diffusion is not the right way to do what we’re doing. There is a company that actually does the stable diffusion version of what we’re doing, but they’ve never shown any proof that it actually works. So, from what we know and how stable diffusion really works, from a research side, we believe that that technology is maybe five to 10 years away. I definitely feel AI should be included in accessibility solutions, but it’s not mandatory.

Where do you see SignAvatar in the next three to four years?

We plan on raising either a pre-seed or seed round, depending on how big the company will be in a couple of months. Our goal is to move management into the US. I plan on moving to San Francisco at the beginning of October, and then slowly but surely, move the business into the US. 

We plan in the next two and a half to three years, to open another vertical of our solution, to go into education or something else that we’ll do market research for. But in the next three to five years, we hope to be in 20 to 50 airports in the world, primarily in the US – that’s our goal.

Subscribe

Login

Notify of

new follow-up comments new replies to my comments

Label

{} [+]

Name*

Email*

Website

Label

{} [+]

Name*

Email*

Website

0 Comments

Newest

Oldest Most Voted

Inline Feedbacks

View all comments