In what place did we work?
Our office has not yet found clear boundaries. Verigram often moved from one location to another. If there were no empty locations, then the next one was chosen, which is close to the previous one.Sometimes I had to work hard to find a new office. It was especially difficult on weekends.
Why did the Verigram team move so often? - you ask. The answer is very simple - our office was a cafe. We looked for and gathered in such a location, which was convenient for everyone to come.
The relocation continued for three months until we settled not far from the Botanical Garden of Almaty in a place called "Malina Mix". I even had a loyalty card, although we ordered only a teapot. And yes, the waiters disliked us. This is clear - we occupied the table and ordered almost nothing.
At first, we met once every 1-2 weeks. There were breaks and it was not possible to see the guys for 2-3 weeks. We held status meetings with elements of short brainstorming.
Operational work and execution took place remotely from home.
What did we do in the meetings?
We did, what freshman entrepreneurs do. We were searching for the Problem.
The founders' previous experience clearly indicated that no company will survive without solving the actual Problem. CBInsights statistics also confirmed this: "42% of startups die because founders are trying to solve interesting, but irrelevant problems." The only thing we knew was that we wanted to build a product in the field of computer vision and use artificial neural networks. The area is vast, there are many tasks for implementation. But how to find a Problem out of a thousand? In order to concentrate on working on it, expand and deepen your knowledge. And most importantly, to create a Product out of this Problem.
What was the first hypothesis?
In 2017, facial recognition technology was on hype. All major publications in the USA and Europe wrote about this. The fire spread to the CIS. At least one article a week was published that talked about the wonderful technology and the advantages of using it.
To be honest, we got under this hype a bit. No matter how we tried to bypass this direction in search of something new, for some reason we were tempted to deal with the issue of remote identification in Kazakhstan. In the end, we gave up.
We started to dig up the problem of facial recognition: in terms of business, we prepared a Business Model Canvas and evaluated the Kazakhstan market by indirect methods; from a technical point of view, the guys delved into reading scientific articles and searching for a ready-made academic model to build the first prototype.
After a few meetings and a couple of months, we were able to summarize the first results:
- Technologically, the project is built on convolutional neural networks and uses deep learning technologies. But there is one problem - there is no ready-made dataset for training the artificial neural network.
- The business model did not hold water. Meetings with the first potential clients spoke of a reluctance to conduct pilots and use facial recognition technology in business processes (the situation has changed a lot in 2020).
For what did we get paid for the first time?
By chance, we received an order for text recognition. The client wanted to recognize vehicle license plates from pictures. We were in no hurry to make a decision.
From the engineering perspective, the order was great:
- it is computer vision,
- work was related to proven technologies for text recognition,
- experiments with neural networks and machine learning will be required.
To decide on the fate of the order, a vote was held. We chose for a long time, in two iterations, and thought even longer. In the end, we agreed to help the client.
It was decided that while the technical team was gaining experience in Optical character recognition (OCR) and machine learning, the business team would continue to look for the Problem.