Title goes here

About us
Total 39 companies invested
Total 79 companies invested
Today we talked to Leo Aleksandrov, the founder of Brickit. Brickit is an app that assists in playing with lego bricks, providing ideas for constructions and helping to find pieces in a heap.

Located in the United States, originally founded in Russia, Brickit transcends geographical borders. This startup has over 400,000 MAU, with the United States, United Kingdom, and Germany as its top markets. This year Brickit has achieved an ARR of $1,000,000.

We asked Leo how the idea originated, what difficulties he faced and what Brickit has in store for the future. Here is what he shared with us:

Brickit is a creative assistant for playing with lego bricks

Brickit is an app that comes up with hundreds of ideas of what can be built with the bricks you have and shows the exact location of each piece in the pile you’ll need. It's all possible thanks to the wonder of machine learning.

Brickit encourages kids to rediscover the joys of playing with bricks, stimulating their imaginations and creativity to the fullest. It saves parents from having to search for stray pieces on their hands and knees as well.

Furthermore, Brickit provides tools for creators to showcase their work and receive the recognition they deserve.

How we got our start

It all began for Brickit in 2018 when my young son needed stimulating outlets for learning. We bought him the usual toys, but they soon became dull and unappealing. That's when I remembered LEGO; with its boundless creative potential, it was the perfect solution.

Soon, I was buying tons of LEGO sets for my son, and all the pieces accumulated into one massive heap. Finding the specialized parts and figuring out builds with the most suitable combination of pieces became a great challenge. Let me tell you, searching for an hour and a half just to find the right piece to build an elephant is no easy task!

By 2018 I had already attained the position of product manager at Yandex, a leading Russian IT company. Knowing this, I instinctively approached the situation from a business and product angle. I began to discuss my experiences with other parents, both in my life and online forums, in order to uncover if anyone else had encountered a similar issue. To my great surprise, many individuals were not only feeling the same pain but were willing to pay to see it resolved.

At the same time, there was a phenomenal surge in computer vision technology. Subsequently, I proposed to my Yandex associates that we develop a neural network-based solution to accurately find the pieces of a LEGO set among the disassembled parts. Surprisingly, they were interested and agreed to join me in developing such a product.

In the beginning, it was just the three of us: Andrey Tatarinov, our CTO and co-founder; Evgeny Mozharovsky, our first mobile engineer; and I. Now we are a 16 people tech-savvy team with a working background in tech giants like Google and Yandex spread across Europe and the US.

Difficulties we faced

My colleagues at Yandex told me that the solution we needed was going to be straightforward, but that wasn't the case. We really had to struggle.
Developing an MVP for a product related to neural networks is expensive. Training a neural network takes a long time and is really costly.

However, I wanted to validate my hypothesis, so I decided to put it to the test with the help of my son. What I did is put the LEGO parts by color into different boxes. Amazingly, it reignited a child's enthusiasm for playing with Lego.

Then we created a straightforward landing page offering to sort LEGOs for £4 per kilogram in just one day. Once we started driving traffic to it, orders came flooding in. I should say, sorting through the 50kg pile by hand was admittedly quite a challenge…

All in all, we've tested our idea with the least amount of resources.

Obtaining test samples was quite a challenge. A neural network will not work properly without a good dataset. And sadly a suitable dataset with heaps of LEGOs cannot be found on the Internet.

So what we did first was a service wherein you choose the LEGO sets you've purchased and the app suggests what things you can build from the parts you have in the collection.

​​At the top of the app, we included a bar inviting users to take a photo of their LEGO pile. We informed them that doing this would help train the neural network, and they happily obliged! That's how we got our test samples.

On launch day we experienced server crashes due to the overwhelming response of 40,000 users taking advantage of our Finder mode application which utilizes neural networks.

We were ready for explosive growth, having seen a friend suffer the consequences of unstable servers when his service saw high demand. Anticipating this, we made the decision to shift the majority of the calculations to users' devices. Unfortunately, this meant the app first  ran only  on the most high-end iPhones, and not always optimally.

Despite the majority of the calculations being done on the device, our servers crashed twice. Thankfully, our engineers worked heroically and managed to fix them quickly, and there weren't any major issues.

What's next

We have a ton of plans for new modes and collaboration projects.

  1. We've been collaborating with Snap and we'll be releasing a demo version of Brickit in their app this spring. Snap will have more content, and we'll get more of an audience.
  2. We're plugging in generative neural networks to generate content for a media feed. The media feed provides a platform for creators to showcase their buildings. Neural networks will add the AI-generated surrounding and a narrative to the pictures.
  3. Finally, we're developing a new mode which is in secret now. “

We are glad to have Brickit on board, helping them to enter new markets and follow their new perspectives. Excited to see the next steps guys are going to take.

Join us, stay tuned, and follow Brickit in their social media:
LinkedIn, Instagram, TikTok