25 Mar Clever interviews: Machine Learning genius!
What is Clever? We ask our Machine Learning genius, Fernando Constantino
What made you start your own business?
I, along with my partners Manuel Climent and Rodrigo Muñoz, thought there was a business opportunity (there were just low quality apps to build adwords campaigns or they were too expensive) and we had the technical expertise to build it.
You studied business, how did you end up in the world of Machine Learning?
I also knew how to code. To me, ML seems like the shortest path to finding a solution to many predictive needs we face in business. I tend to be very competitive, so I started in ML by taking part in ML competitions.
In summary, what is Machine Learning exactly?
Although the ML field has grown to be very complex lately, I’d still summarize it as a way to make a machine learn automatically through feeding it examples, without having to explicitly code any lines of code.
Why is there so much (growing) interest in machine learning and big data lately? What advantages do they have for companies?
You can see ML (specially deep learning) as an automatic universal approximator to almost any predictive business needs (among many other needs). Now there is no need for long costly programming/statistics projects; businesses can now cheaply, quickly & accurately manage many of their needs, and even go the extra mile, providing extra services through chatbots and other ML powered automation.
It is undeniable that there is also a little bit of hype, which will eventually fade.
If you could use Machine Learning for some kind of personal project … What would it be?
It would be nice to build an automated score system for my tennis matches, using a live stream from my cell phone.
What is the last thing your team developed?
The capability to automatically build adwords campaigns for any site, without the need of any structured feed.
Which of your projects are you the most proud of?
Although predictions tasks are quite profitable for Clever, I am usually happier working on projects that provide some degree of automation, so people in Clever (and our Clients) can focus on real value-adding tasks.
What are the current challenges of Machine Learning? And what do you think they will be in the future?
Some ML tasks, especially Deep Learning, are very resource intensive. This can be solved through better algorithms or cheaper hardware. I think it’s pretty obvious that the future will be filled with APIs, requiring zero technical knowledge. Most companies won’t be training models themselves, or if they do, they will start with pretrained models.
Can you briefly explain what Tensorflow and Spark are?
Tensorflow is mostly known as a deep learning library. It’s slightly complex, so many people use it through simpler libraries such as Keras. Spark is a well known tool for handling Big Data. My focus is more in ML than Big Data.
How do you apply Machine Learning to the ecommerce environment?
There are many ecommerce tasks that benefit from ML which are being used in Clever, such as visual/language categorization, taxonomy matching, conversion likelihood, chatbots, client ranking and churn prediction.
A tip for boys and girls who want to be future experts in Machine Learning
Start with a Mooc (i.e. Coursera). In parallel, take part in a Kaggle ML competition. First try to get a bronze metal in Kaggle, then a silver medal, and so on… They should also read this article about what the future holds for e-commerce.
If you had to define Clever with 3 words … Which three would you choose?
Automated Adwords Creation.
If you want to read more interviews like this to members of our team, do not miss our posts, subscribe to our newsletter, and follow us on social media!