Claudi’s workday: Machine Learning Specialist – YA

Claudi works via Young Analytics as a Machine Learning Specialist for an international bank in Amsterdam. We were wondering: what is it like to be a Machine Learning Specialist?


Claudi, explain us what it is exactly what you do

As a Machine Learning Specialist my job is basically solving issues by designing and implementing intelligent systems. For example, I created a model for the Risk Management team that can find anomalies in big financial datasets by itself without having explicit examples of them. To accomplish such cases, my daily activities are a combination of programming and reading scientific papers. I have to make things work and that’s what I think is challenging about my job!


What does your regular workday look like?

First of all, my job never ends. When I go home I still think about what I have done that day and what I’m going to do the next day (curiously, most of my good ideas come up when I’m taking a shower). Normally when I have been assigned to work on a new project, I first have to do research in the state of the art of machine learning (one of the approaches to artificial intelligence). When I find a technique that can be used to approach the problem, I implement it and I evaluate the results. The models I have built constantly improve themselves overtime because they get new data every day (the anomaly detection model for instance, gets around 10 million samples a day). So in short I’m fighting with the machines on a daily basis!


What is it what you like most about the world of Big Data and Machine Learning?

I like that there isn’t much done out there. You don’t have many sources of knowledge of “how to” in the machine learning field or open datasets in every domain (like open financial datasets), so you basically start from scratch. Sometimes you can contact someone who did something similar, but most of the time you have to find the answer yourself. It’s like doing a PhD but without having to publish. One way or another: it has to work!


What kind of program tools did you learn or did you already know?

I didn’t have the skills and knowledge I have now when I came in. When I joined the current bank I knew Python and had knowledge in supervised machine learning. Now I know more about Hadoop, Spark, TensorFlow, parallel computing, hyper-parameter optimization and unsupervised machine learning. What I like about my job is that I learn something new every single day, and that I walk out the office with the feeling: yes! I learned something new today!


What was your biggest challenge so far?

The first months at the company were very stressful because I didn’t have a GPU and without it I could get the output from a model only after two weeks. These weeks I could do nothing other than reading scientific papers. Now, with GPUs I already know the next day what a model’s output is! Everyday is a challenge, because I have to prove that the model or algorithm I built is the best solution for the specific need!


What are your ambitions?

I told my manager few months ago that in the near future, we will have rooms full of GPUs (Graphic  Processing Units) training our models. You can compute some operations by using CPU’s, but GPU’s are much faster! My dream is to design models that have a big positive impact within the company and hopefully for society as a whole!