What you get:
10 ideas for machine learning projects.
Free datasets to use for your projects.
Examples of libraries and algorithms for each case.
It took me two years to write these words, but I think it is about time I share my experience. Let's start assuming my responsibility in the bankruptcy of Samsamia, I was the CEO, and I, not my cofounder and ex-best friend Miguel Angel Maldonado Agramonte, was to blame. It was my poor choices and lack of future vision that made a company with around €200k in revenue and a bright future collapse.
The idea of Samsamia 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 walking in Madrid 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 at half of the stores in Madrid, but it was impossible to find. She was so desperate that she sent me pictures 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 them: "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 thought well, I won't be able to find the bag, but I could use machine learning and computer vision to do it. Samsamia idea was born.
By then I was a PhD student in robotics, so I started to code the first version of the search engine using OpenCV and C++ (later this code was translated to Python).
I knew how to create an algorithm that extracted visual features of a garment and compare these features with a big database of other garments.
However, I had no idea how to make a phone app or an API. So I started looking for partners.
After being rejected by several friends and colleagues at my research lab, I asked Miguel Angel Maldonado to join me. I had been his Master Thesis director and he had experience in developing server-side applications. We founded Samsamia Technologies with 50%-50% shares each and he became the unique administrator, due to an incompatibility with my current job at my research lab.
I didn't know back then but that was a decision that I was going to regret in the future. I thought that a shareholders agreement can protect my interests as a stakeholder, however, that didn't prevent the events that happened afterward. This mistake cost me the company, I barely knew him, we had worked together for a while, but that's completely different to work with a person under high-pressure conditions and with money entering the company.
The first product Samsamia released was called Dresscovery, it took 1 whole year to develop. It was an app that allowed you to take a picture of a bag and buy a similar one available for sale on online e-commerces.
We made some noise in the Spanish press and won some international contests, but the product was a complete failure. When starting the company, I had no idea about the concepts of product development vs customer development, the lean startup method, or growth hacking.
I did the rookie mistake of many technical entrepreneurs: have a technology, not a product. When the app was released to the public, there was no engagement and it almost made no money. It was the end of 2013 and we had a product that nobody wanted.
The year 2014 was really tough with almost no money in the bank. We learned from Dresscovery mistake and pivoted from B2C to B2B. I was spending most of my time in finding customers to sell them the visual recommender through an API.
Meanwhile, Maldonado keep complaining about the difficult situation of the company and showed intentions to leave, but I felt that we had a big opportunity waiting for us. We were running out of cash so we decided to ask for money from our families. During this year I put €8400 in the company, which mostly went to his salary. But when I asked my partner to match that quantity he didn't do it. We agreed that I had to add the money as a loan, instead of changing the percentage of the shares between us. We also agreed that the company had to reimburse the loan the moment we start having cash. However, that never happened.
At the end of 2014, I was able to land a really big customer. They were one of the biggest e-commerce in the UK and money started flowing to Samsamia. I was so excited, my effort and sacrifices had paid off. The visual recommender that I initially created and then we evolved in the company, was going to face real customers.
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 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's need. I asked a manager from our customer once: why did you choose Samsamia, a small startup in Spain, instead of this other competitor that had £4M of investment? They replied that the other competitor had shown 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. Samsamia, in contrast, was willing to modify the algorithm to fit the customer's requirements.
Back in Spain, the situation with my partner was getting worse. Money was flowing in the company and my partner was not hiding at all that he wanted me out. First, Maldonado created an "advisory board" with a friend of him, Jaime Rosco, who by then was working in the company, the accountant (which was a friend of his dad) and his dad (yes, his dad, who was a professional seller of boilers from Bosch). He also wanted more equity and control and asked me to generously give part of my equity to his friend Rosco. Sometime later, Maldonado and Rosco flew to London without telling me (and paid by the company) to try to convince one of the customer's managers to take my position as CEO (the guy freaked out, by the way) and they started developing a parallel version of the visual search code in a repository that I had no access.
Around the end of 2015, everything became crazy. Using his position as administrator, he removed my access to the company's bank account, I was not able to see what Maldonado was doing with all money the company had. Not only that, after my lawyer claimed my debt back, he stated that the company had a debt with him of a little bit less than €20.000, money that never entered the company, part of which was from even before the company was incorporated. I rejected that debt, and as you can imagine, I never had my money back.
Finally, in early 2016 Samsamia stopped paying me. The moment that happened, our main customer stopped paying the company. From the €70k that the company had in the bank at the end of 2015, most of that money was gone by summer. As an investor and owner of the company, one of the main rights is access to information, however, I don't know what software development or sales initiatives the company did between January and September 2016. After the summer, my partner and the other workers in Samsamia had to find other jobs. The company was destroyed and I sued Maldonado.
As you can imagine, I had a very hard time in all this process. But guess what, making a startup is hard. Surprise! I reflected a lot about all those years.
To make a startup is brutal, I have worked in science and then in one of the top tech companies in the world (Microsoft) but there is nothing as hard as creating your own business. The pressure that you experience is extreme. Therefore, the behavior of people under the circumstances of a startup is difficult to predict. If you are trying to start a startup with a friend, think about this, do you really know this person? are you confident that under high levels of pressure this person is going to help you or is he going to break apart? It is super common to divide shares generously on the first day of the startup, it's like marrying a girl the first day you know her. Isn't it creepy?
I think that the key to all my success and failures in Samsamia is adaptation. I failed when I was not able to predict my partner's behavior or when it took us so much time to develop an app that nobody wanted. I succeeded when we were able to adapt to the customer needs, like when we changed from B2C to B2B or when we adapted our technology for the customer.
I guess the biggest takeaway for me is the feeling of accomplishment. A small company in Spain, with no investment and internal problems, was able to create a product that one of the biggest e-commerces in the world used. It was Reid Hoffman who said that making a startup is like jumping from a cliff and trying to assemble an airplane on your way down. I have the feeling that at least I built a crappy plane that flew for some time and then it crashed with a huge explosion. The time that the plane flew was amazing.
Here is a demo of the product we made for ASOS.