Carlos Castro

MIDS Student


Improve the lives of millions of people around the globe, by leveraging data and machine learning at Skype in order to build highly intelligent communication systems.


My first interaction with computer science was writing a QBASIC program when I was 8 years old. I was amazed with the fact that I could create my own games, with my own rules and characters. When it was time to start high school, my ever-growing eagerness towards computing inspired me to switch to a computer-oriented school, where I focused on electronics and programming. This new school was the catalyst for an even deeper connection with science – I was championed to participate in mathematics and physics Olympiads. I had found a new game. Among my many achievements during this period, the most remarkable ones might be winning the first price in the state physics Olympiad 5 times in a row, and being champion at the Argentine National Physics Olympiad. I earned the great honor of being part of the Argentine National Physics team and of successfully representing my country at international physics competitions. As an appreciation for my achievements on behalf of the country, I was awarded a full scholarship for college at one of the most prestigious computer science universities nationwide.

During my college years, I was naturally attracted to the mathematics and computer science courses. I even had the pleasure of having multiple experiences related to machine learning and artificial intelligence, including areas such as Genetic Algorithms, Neural Networks and Anomaly Detection and Optimization.

My most relevant contribution at college before my thesis was in the field of graph theory and network topology. I attended an elective course about internet topology, which was imparted by one of the America’s most recognized complex network researchers. After completing the course, I kept working with him building a highly optimized networks and advanced graph calculations library in C++, which is currently being used by several research teams at the University of Buenos Aires.

After 5 years on college, the last step to complete my graduate studies was, as it is a tradition in Argentina, to complete my thesis work. I was seeking a thesis topic that was special - something that was complex enough to be extremely fun, but that could also show how computer science can improve people’s lives. Thanks to serendipity, I met the president of a non-profit bioinformatics organization, who would wind up being my thesis mentor.

I decided to focus my thesis in bioinformatics, particularly in taxonomy and biodynamics. My thesis had as a goal to integrate phylogenetic data about diseases with geographical data, in order to find the origin and dispersion dynamics of emerging diseases. The thesis particularly proposed an optimality criterion to perform this integration, which was based on phylogenetic tree optimization and tree forest consensus. I presented the thesis in many Biology conventions in Argentina and Chile, getting actionable feedback and excellent acceptance – even potential for future PhD students continuing this work.

An interesting side effect of the thesis was a computer science aspect. Implementing the algorithms created during the thesis’ research phase, I created a highly optimized C++ phylogenetic library which is still being used by a handful of bioinformatics organizations. During the technical development, I published an article in the technical C++ magazine called Overload, in which eminences such as Bjarne Stroustrup – one of the creators of the C++ programming language - usually publish. The article was titled ‘Curiously recursive template problems with Aspect Oriented Programming’, and had extremely positive response from the C++ community.

After graduating from college, I had my first startup experience – I joined the team at Kormox, a security and data flow distributed systems startup based on New York. I spent almost a year there, until I got a proposal that would change the course of my life: I was interviewed and instantly made an offer from Skype, where I currently work as a Software Engineer.

It has been 3 years since I joined Skype, and the depth and breadth of my learnings is incommensurable, and I can say I finally found a job that is, at the same time, my hobby. At Skype, everything’s scale is humongous. There are more than a billion Skype accounts, with the number of users greatly surpassing the 300 million. On average, the total number of skype calls per day is beyond 3 billion, and since the beginnings of the company, the cumulative minutes of calls is about to reach 2 trillion minutes. There is a quote often stated by computer scientists, which states that even if you solve a problem for a certain amount of information, solving it for an amount of information which is 2 orders of magnitude higher is a completely different problem. That is the case for my day to day work.

The different contributions I’ve achieved are of great relevance – I was among the select group of engineers who decided to re-architect and re-write Skype core, moving from a monolithic scheme to a cloud distributed service-based architecture. I personally wrote the storage and configuration systems of this new architecture, covering the diverse aspects that this encompasses – Fault tolerance, high availability, disaster recovery, caching, pre-fetching, compliance, security and several more.

During my career and particularly while working at Skype, I’ve seen countless examples of the power of information, and particularly of the value that extracting insightful data from huge piles of random information can add. Not only is the business impact that leveraging data can provide enormous, but it is also thrilling how fun and rewarding the process can be. My experiences with machine learning at scale and analysis of complex data sets have acted as a catalyst for a great new passion in my life. Even though I am, by nature, a computer scientist, I believe that being a Data Scientist in addition would greatly broaden and deepen the impact I can have at Skype or any future endeavor.

One of the reasons why I believe Data Science at Berkeley is the perfect match for my needs, is how multi-disciplinary its curriculum is. For example, at Skype we deal with billions of users – understanding the needs of those billions of users in order to provide a better product does not lie only on computer science, but on social sciences, behavioral sciences, statistics and many more. Additionally, the program also relentlessly focuses on the scale aspect of working with and analyzing data - an angle which I consider mandatory given the petabyte level of data I usually deal with and the imminent ever-increasing data scale for internet applications in the years to come.

Last updated:

August 18, 2018