Search in Progress...

On Campus Jan 27, 2018

A brief coverage of the talk by Mr. Pravin Shankar, Data Science manager at Facebook and Alumnus of MSc. Software Engineering (1998-2003)

Data Science- a topic that has been fervently discussed across the world for the past few months and has stimulated the interest of the current generation. A lecture on the hot topic was something that both the students and faculty were looking forward to. The talk was a part of the commemorative lecture series, organized by the Applied Mathematics and Computational Sciences Department and was attended by the Head of the Department, faculty and eager students.

A card with one’s name followed by ‘Data Science Manager at Facebook’ is a dream for many. Most would hardly expect a simple and down-to-earth personality in a person of such stature, but that is exactly what Mr. Pravin Shankar was. After completing his five-year MSc. Software Engineering course in PSG Tech, he moved to Bangalore for work. In less than a year, he took the plunge, quit his job and chose to pursue a doctorate in Computer Science from Rutgers University, New Jersey, where he specialized in Machine Learning, Neural Networks and Data Analysis (what Data Science was previously known as). He has been working in this field for more than a decade now, even before it was an “in-thing”, as he says.

Continuing, he added, “I think the main cause for the popularity of Data Science is due to the rise of tech giants like Google, Facebook and Amazon, and the fact that they have user bases of billions, which is analogous to more data.”

The speaker, Mr.Pravin Shankar

Moving on to how Data Science has influenced the world, he quotes examples of how we get Netflix recommendations, related posts on Facebook, and Google search results. “Working in a company that has a humongous pile of data, and the feeling when you realize that the code you write is shipped and used by billions of people is so fulfilling.”, he added. The sense of contentment was evident in his words. The crowd let out a gasp, when he showcased the above figures in words. 2 billion people of the 7 billion in the world use Facebook every month (Do remember that the 7 billion includes people under 13 years old and the entire population of China). And adding to the count, services like Whatsapp, Instagram and Facebook Messenger, all of which come under the same roof, also have a huge user base. Elaborating on his previous statement, he displayed a map indicating communication between different parts of the world, which showed the power of modern communications and the role played by Data Science has in it.

Addressing the question “What do Data Scientists do at Facebook?”, he answered “Product Inception”. (The audience were as clueless too). “Generally in any development team, there always exists this question of “What next?”. It is answered either by intuition (read Steve Jobs) or by user suggestions. Taking suggestions manually is impossible and sensible suggestions may go unnoticed in the converse. That is when Data science comes to rescue, which is how the ‘Marketplace’ feature in Facebook was launched. Many big groups existed in Facebook in which people carried out business transactions and trading, which was not anticipated when the groups feature was started initially. This eventually triggered the idea. Similarly, they define the problem, write suitable code, ship it and test it with a smaller set of people and if good, build the required feature and launch it to the user. On having succeeded in the launch with the help of data science, people realised how useful it was and which regions used it the most. Furthermore, they customize it to ensure it adheres to the usage in different parts of the world”, he explained.

Food for thought

Delving deeper into the “Search” option in Facebook, “The hardest problem is inferring to the context in which the user searches for a particular page, and choosing among many other pages with similar titles. For example, there might be hundreds of alumni groups of a particular college. Among them, the search result should display the most relevant group to the person. And so, information on the user’s education can be retrieved from his profile, in order to provide the best results. The bigger role played by data science in Facebook, compared to Google and other search engines, is that Google search displays the same content for everybody, while Facebook search sorts the content with relevance to the user”, he elaborated. He then extended the topic to the steps involved in search, about Indexing, Retrieval, Query understanding and Ranking.

The keen audience

Moving on to measuring the efficiency of a search engine, he explained why the number of users as a metric is not as good as the click rate metric, due to the huge amount of data and time complexity involved in the former than the latter. He then threw some light on “position based click rate”, which is a far more optimized measure than the normal click rate and that it works based on the reciprocal mean rank.

Towards the end of his lecture, he defined three qualities of successful data scientists – Programming ability, Qualitative analysis, and most importantly- Product and business thinking. He also cleared all the queries of the audience regarding search relevance, databases and scales, privacy and assured that user data privacy is taken very seriously.

The Head of the Department, Dr.Nadarajan recalled few nostalgic memories of the speaker and his batch and proceeded to deliver the vote of thanks, which was followed by the presentation of a memento to Mr.Pravin by Mr.Karthik (MSc. Theoretical Computer Science, batch of 2007-2012), who is also a Data Scientist and founder of Probyto. Finally, the crowd departed with an insight into the world of Data Science.


Ananya Haraprasad

Along with Mukund E

A low-budget Princess Carolyn juggling multiple tasks to escape existential crises; found making puppy eyes at dogs or binge-watching series otherwise.