Computer vision technology has taken a massive leap forward in the last decade. Smart cameras are now capable of recognizing objects, enabling their successful application in security systems, autonomous vehicles, smart home ecosystems, and various other fields. The startup Bizdin.ai has taken computer vision to the next level by leveraging artificial intelligence technologies, teaching cameras not only to identify objects but also to understand what they see. For instance, if an AI-powered camera detects a fight in a school hallway, it can instantly recognize the situation and automatically notify the administration about the incident. CEO and founder of Bizdin.ai, Aibek Nurtazayev, spoke to ER10 Media about the capabilities of their Large Vision Model (LVM) technology.
Follow Kazakhstan’s Startup Movement in the "100 Startup Stories of Kazakhstan", a collaborative project by ER10 Media and Astana Hub. This initiative highlights the most innovative Kazakh startups, showcasing projects that stand out for their creativity and impact. Among the heroes are Astana Hub residents, as well as creators of other innovative technological products and services. The content is available in Kazakh, Russian, and English.
AI: Seeing and Understanding
— Can you tell us about the idea behind your startup? What technologies is it based on?
— We are creating AI agents “with eyes” that can see the world almost like a human. In other words, a camera with our AI model not only recognizes objects but also understands what is happening. Moreover, our neural network can describe the processes it observes. This solves two major problems. First, we are democratizing computer vision, enabling anyone to "interact" with cameras and customize them for their needs. For example, imagine you are a school principal. There are hundreds of cameras installed throughout the campus. You can assign tasks to them in plain human language. For instance, you could say or write: “If you see fighting, bullying, or smoking, immediately report the incident to me”. As soon as the algorithm detects any of these activities, it notifies you instantly.
Another example: a camera with an AI model can act as a museum guide. You use your smartphone, point the camera at a painting, and the neural network provides you with fascinating information about the exhibit.
You can also interact with the artificial intelligence model. Let’s say you’re in San Francisco, pointing the camera at the famous Golden Gate Bridge. The neural network instantly gives you all the information about the landmark. Or, if you’re in a supermarket, you can point the camera at a shelf and ask it to identify only vegetarian or halal products. The neural network can easily do this in real time.
— Can this AI model be installed on any device with a camera?
— Yes, the algorithm can be connected to surveillance cameras or installed on a smartphone as a helpful AI assistant. Additionally, the AI algorithm can analyze video footage or photographs. Currently, our LVM far exceeds the capabilities of standard computer vision.
A Technology in Demand
— Is the technology already being used anywhere?
— Yes, we are currently finalizing a contract with Kazakhstani startup Parqour to provide our technology for safe parking. We are also running a pilot project with NARXOZ University. There, our cameras notify parents when their child or student has safely arrived at school or university. Additionally, we provide analytics: if the algorithm detects that a student has been arriving in a bad mood for several consecutive days, it alerts the school psychologist. The cameras also monitor for order, reporting any incidents immediately to the administration.
We are also collaborating with a team developing glasses for visually impaired people using our technology. The gadget helps users navigate their surroundings by providing verbal descriptions of obstacles or the environment.
We are conducting pilot projects in Hong Kong, Singapore, the Philippines and Malaysia and are gradually entering the US market.
There is also a project with Samruk-Energo JSC at a strategic facility. Certain zones are restricted to authorized personnel only. Using our technology, we’ve implemented an identification and access control system. Importantly, we only provide the technology and do not have access to their data.
— So, does this mean your technology is best suited for security purposes?
— Not exactly. Our technology can be applied to analyze any type of video, photo or image. For instance, we are starting a partnership with Getty Images in the U.S. to optimize image searches. Imagine you need an image for an advertising banner featuring a cat and dog dancing while being filmed by a cameraman in a red jacket. Using a standard search engine could take a lot of time because conventional engines, designed for SEO and tag-based searches, don’t understand subtle nuances. Our algorithm, on the other hand, comprehends the context and meaning of the query, allowing it to quickly analyze and deliver the needed image. Essentially, it’s like ChatGPT but focused on videos and images.
An Unconventional Success Story
— How did the idea for your startup come about? Was it inspired by market demand or your own experience?
— Our history and sequence of events is completely unconventional. Typically, people conduct customer development, identify a problem, and come up with a solution. For us, things unfolded differently. Our team has been working in the AI field for six years, with several products already under our belt, such as Shyraq.ai and Sellstat.ai. However, we wanted to create a truly disruptive technology that would shake the market. We realized that most scientific breakthroughs in technology appear roughly a decade before their commercial adoption. Assessing our capabilities, we identified three potential directions for growth: AI, quantum computing, and space technologies. After considering various ideas, we decided to focus on AI technologies.
At that time, OpenAI had just announced ChatGPT, and we decided to launch a platform where clients could create their own AI bots. We succeeded in developing the platform, but within a month, OpenAI introduced its own platform, which we couldn’t compete with. Then we launched a marketplace for selling chatbots. Sales began, but within another month, OpenAI launched a similar marketplace. It became clear that whatever we did in the realm of large language models (LLM), OpenAI would eventually enter the segment and outcompete us.
We decided to focus on a niche that has its own specific challenges but no competition from the big players. Typically, if a small player solves a market problem in a particular segment, large companies prefer to acquire the competitor rather than enter the market themselves. That’s when we chose to concentrate on video analytics.
At that point, we got lucky. We exited one of our companies, which gave us some initial capital. Around the same time, cryptocurrency mining, especially for Ethereum, was highly popular in Kazakhstan. When Ethereum transitioned from Proof of Work to Proof of Stake, all the GPUs used for mining became obsolete. This allowed us to buy a large volume of GPUs at five times below market price. We thought, “Great! We’ve got a ton of GPUs — let’s use them for something, like training neural networks!”
— At what stage is your product now? It seems to be operational already.
— Yes, the product is fully functional. We’ve developed an MVP, and now we are launching a platform that enables anyone to quickly adapt the AI model to their specific needs.
— Do you currently need investments to develop the startup further?
— Overall, yes. Until now, we have been financing the project on our own. Now, we’re seeking $5 million in investments. While we could continue developing without these funds, such an investment would allow us to grow much faster. We’ve already received offers, such as a $1 million offer in Kazakhstan and an $8 million offer in Southeast Asia for 40%-50% of the startup. However, selling such a large stake isn’t justifiable when looking at long-term goals and aspiring to become a unicorn.
— So, you’re planning to become a major global unicorn?
— Absolutely! This is our focus. I’d also like to note that our team members are more researchers than developers. We’ve authored several scientific papers. AI technologies evolve rapidly, and maintaining leadership requires a strong emphasis on R&D. Additionally, we aim to create a virtual studio where companies and clients could train AI models for their specific needs, exchange experiences, data, and best practices. Such a studio could bring tremendous value to the field.
— Do you have competitors?
— Yes, we divide them into two groups. The first includes companies that use LVM technologies (e.g., Google and OpenAI), while the second focuses on security. We differentiate ourselves from the first group by ensuring that we never transfer data to third parties and can organize the entire system locally, even in areas without internet access. Compared to the second group, our algorithms understand context, and our solution can be adapted to any industry.
A Global Mission
— What mission do you see in your startup? Are you aiming to fulfill a particular dream through this project?
— We want artificial intelligence to solve real-world problems. Today, we live in a world driven by data. Neural networks perform most effectively when trained on high-quality data. Our goal is to teach AI to see like a human and understand its surroundings. With this capability, AI will become highly useful and solve numerous tasks. AI-enabled cameras could prevent crimes and disasters, assist in research, and perform many valuable functions.
— Have you attended any training programs or accelerators?
— We had enough experience to start without accelerators, but as we expand, we are considering American programs such as YC, HF0, and others.
— How long have you been in business? Have you ever worked as an employee?
— I’ve been in business since 2017. I became an entrepreneur right after graduating. I’m also currently attending classes at Stanford.
— Do you enjoy being an entrepreneur?
— I love being not just an entrepreneur, but an engineer-entrepreneur. It’s like Elon Musk or Steve Jobs — when you’re not just growing a business but creating new technologies and essentially being an innovator. I want to create something truly important and meaningful for people.
— Do you want to contribute to the development of Kazakhstan?
— Absolutely! I am a member of Qaz AI, the Association for the Development of Artificial Intelligence. We strive to promote the development of artificial intelligence in Kazakhstan and actively support startups in this field.
I am also a member of the Expert Club under the Senate of the Parliament of the Republic of Kazakhstan, where we assist legislators in understanding AI-related issues. We plan to support local startups to help them enter international markets and bring world-class experts to Kazakhstan.
Additionally, we developed the Shyraq.ai project, which helps teachers prepare for lessons without spending excessive time on monotonous tasks.
As you can see, our team is committed to contributing positively to Kazakhstan.
— What inspires you in business? Do you have any favorite books or movies?
— Honestly, I don’t have time for fiction. As for movies, The Men Who Built America is at the top of my list. I also love films about top entrepreneurs. I frequently rewatch series like Billions, Suits, and Silicon Valley. I can watch them 10-15 times, not for entertainment, but as a learning course. If you pay attention to the details, there’s a lot of valuable information. Apart from that, I draw inspiration from people. I enjoy meeting and interacting with others.
— What sport would you compare to your business approach?
— Sailing! It’s about freedom, limitless spaces, and beauty. A yacht club called Go Sail is opening soon in Kazakhstan. As a member of this club, I would like to contribute to the development of this sport in Kazakhstan.
