RecoMind

We decided to create a new brand: RecoMind, it is a powerful name whose relationships with recommendation systems is quite clear.

We also decided to make a big change in the business model, because we charged based on recommendation requests, all our competitors follow the same approach. However, you could recommend a million products and nobody would buy it. That was not fair for our customer.

Our main objective is to make our customers successful. Therefore, we decided to change the business model. From now on, we won't charge any upfront fee or cost per request. We will only charge a fee when one of our recommended items is sold.

A new chapter in the story begins.

We made a project with an automotive company. The customer wanted to explore a number of research questions in the context of recommendation systems, such as what kind of recommender algorithm is most adequate to the data; does the inclusion of user or item metadata improve the performance of the recommender or how the different car options affect the model.

Our team developed a number of content-based filtering, collaborative filtering algorithms and other machine learning assets, such as cooccurrence features, text embeddings and performed feature importance analysis.

At the end of the 2014 we were able to land a really big customer. They were one of the biggest e-commerces in the UK and money started flowing into the company.

During the time we were working with the e-commerce, I learned a lot: how to create mission critical workloads, what are the challenges of a big e-commerce or how different is to have a visual search algorithm than to have a product.

One of the top learnings I got is how important is to have a product that covers the customer need. I asked a manager from our customer once: why did you choose us, a small startup, instead of this other competitor that had £4M of investment? They replied that the other competitor had showed their visual search results and they asked them to modify the algorithm to better fit what the e-commerce users wanted to see. But the competitor didn't do it, because they wanted a general tool. In contrast, we were willing to modify the algorithm to fit the customer's requirements. That's how we won the contract.

We incorporated the original company.

We wanted to be a new communication channel between fashion brands and customers. We developed a search engine for fashion products that allows a smartphone user to find whatever he or she desires just by taking a picture and buy it instantaneously. Our application aimed to be the future driver of online retail growth through the connection of the online retail business with the real world.

Technologically, the application matches any picture of an apparel or accessory with a large database of online brands stock. It allows the user to buy the targeted object in the respective online store. The app finds the most similar products that are available for sale right now among hundreds of different brands.

The idea came to me when I was jogging near my home. The previous day, my sister had told me how frustrated she was because she couldn't find a bag that she liked. She was taking a walk and saw a girl with this particular bag. She stopped the girl and asked her where she had bought the bag, but the girl said she didn't remember.

My sister looked for the bag online, she went shopping to half of the stores in the city, but it was impossible to find. She was so desperate that she sent me picture of it and told me: "if by any chance you find this, buy it for me".

Those who know me can imagine how hilarious that was. I hate to go shopping and had no interest in fashion at all. So, I
said to her: "look sis, if you are not able to find that bag, it would be impossible for me to do it".

While I was jogging the next day, I though well, I won't be able to find the bag, but I could use machine learning and computer vision to do it.

That's how I though of building a fashion search engine.