My journey so far

Adriano Marques September 17, 2020

My journey at Exponential Ventures started mid-2019 when I flew 7000 km to Exponential Venture’s headquarters in Chicago. During my first day of work at the HQ, I met the team members, saw their data center with several servers and XNV’s (that’s how we abbreviate the company’s name) lab where there is a laser cutter, 3D printers, and a CNC machine. Right at the beginning two of my beliefs changed: Machine Learning (ML) is easy and cheap and Python has a weird syntax that I will hate it forever.

Before joining XNV, I was working at a Fintech company where I often heard the risk analysts talk about the credit score model or fraud model. Also, they mentioned there was a need to train their Machine Learning model and that it may take an hour or two. At that time, I thought ML was an easy to learn and cheap to train form of Artificial Intelligence. However, I did not know that training some models may require lots of resources such as storage, GPU power, and that, more often than not, it takes days to train it.

Back in 2007, I developed a college project using Python 2.4 to collect stocks data via API, persisting the data to a PostgreSQL database and displaying some graphs in a web application using the TurboGears framework. The professors doubted I could deliver the project because I was using a programming language that most people did not know about and for which there were not many tutorials in Portuguese. Moreover, as I hated Python 2 syntax, I decided to not invest my time learning it after I delivered that project. Years later, coming from Ruby on Rails, learning Python 3.0 and Django was not a big deal. Plus, Python syntax became so much better. In that way, I started to enjoy programming in Python.

Not being a Data Scientist nor knowing anything in the field made my start at the company a baptism of fire, as I had a hard time to understand the jargons and the field itself. I’ve helped to develop two tools meant to be used when doing ML experiments: Stripping and Aurum. Stripping is a tool that speeds up the experiment especially when it will be executed multiple times. Aurum keeps track of all experiments automatically for the Engineer. It will record the parameters, dependencies, metrics, performance, source code, etc. Making it much easier to navigate between the experiments and the experiments’ versions.

To conclude, my journey at Exponential Ventures had a difficult start where many times I doubt myself. So, far I have learned new ideas, techniques, and technologies. Even after expanding my knowledge I feel that the journey is just beginning.

Stay In The Loop!

Subscribe to receive Artificial Intelligence content that will help you perform better.
Don't worry, we don't SPAM. Ever.