Mohamed Maskani Filali grew up in Morocco and attended Télécom ParisTech, which is one of the top engineering schools in France. He enrolled in the master’s degree program at the Data Science Institute at Columbia University in 2017 and focused on machine learning, recommender systems, and user behavior models. Today, he speaks six languages, knows six programming languages, and works as a machine learning engineer at Dailymotion, a global video-sharing platform based in Paris. Here, the 2019 alumnus discusses his work and offers advice to our current students.

Tell us about your current role at Dailymotion.

I joined Dailymotion’s adtech team as a machine learning engineer. We work on subjects related to digital advertising, an exciting and challenging field that offers a lot of room to develop machine learning-based applications. Some of the projects I’ve been working on, for instance, are related to real-time bid prediction at scale (predicting the price of ads that are shown to users in real-time), audience lookalike modeling (segmenting users to allow better targeting), and spam detection.

Which data science skills do you use on a daily basis?

I would say that being able to translate the business needs of the company into a product and coming up with its technical specifications are very important. But on a daily basis, I use all the skills we all learn in the master’s program—algorithmics, machine learning models, feature engineering, as well as techniques for validating and iterating on a model and statistics. I may not use all the tools on a daily basis, but it is good to have a bag of data science tools at my disposal. But no matter what you learn, there is always a need to be creative to make things work in the real world and bring a competitive edge to your company.

What are the most enjoyable aspects of your job?

I am continuously learning and developing my technical skills while collaborating with extremely skilled people. For the past year, I have been deeply involved in the creation of a real-time model serving application programming interface that was integrated in the core architecture of our product. Thanks to this hands-on experience, I’m now able to design end-to-end machine learning pipelines. Deploying machine learning models and applications in production is a fundamental skill to learn today and is a whole world on its own. Also, what I like here is that management has confidence in its employees and encourages you to be as autonomous as possible and bring new ideas to either build new products or rethink existing ones.

How did your DSI degree help you secure this role?

Most of the courses I had were essential to understanding and analyzing machine learning and deep learning algorithms. Using off-the-shelf Python libraries without understanding how they work is useless and limiting because most of the applications you’ll be dealing with will require you to do some customization to fit your needs. I particularly enjoyed the Distributed Systems course, which introduces the fundamentals of distributed systems (compute and databases). It explains how distributing the computation and storage of your applications is essential in today’s data-intensive world. It also gives you hands-on experience with some cloud-based services that are very much needed in the tech industry.

Do you have any advice for current data science students?

I would say that unless you want to pursue a career path in academia, in which case you want to focus more on the theoretical sides of machine learning and deep learning, try to spend your time learning fundamental computer science skills. That will definitely be valuable when you start your first data science job at a company. Place an emphasis on writing clean code and keep in mind scalability issues when designing algorithms and implementing them. Also, explore tools such as Airflow, Docker, and Kubeflow. Why? Because I deeply believe that the most valuable skill of a machine learning engineer is to be able to take care of the productionization of the models you develop, rather than just running them locally. I think DSI students should take advantage of this time spent at home to learn new skills and tools (a lot of online classes and workshops are available), contribute to some open-source projects (think of scikit-learn, TensorFlow), and start some projects on their own to refine their machine learning portfolio. When the job market resumes its strength, they’ll have such top-notch profiles that companies won’t hesitate to hire them.

— Robert Florida