Hi, my name is Vikas Raunak. I am orginally from Madhubani (in Bihar, India), which is quite famous for its paintings. I did my B.Tech. in computer science and engineering at Indian Institute of Technology, Indore. My interest in computer science initially developed by trying to understand how the Internet works. After making a few web and mobile applications, I worked on Multipath TCP as my B.Tech Project, where I implemented & improved upon the key algorithms within the protocol and learnt the complexities of algorithm design in general. However, after reading the paper on TCP Ex Machina, I became more than convinced that learning algorithms from data is the right way to approach several problems and hand crafted algorithms can only take us so far. Incidentally, a lot of people already had similar thoughts (pragmatically coping with a problem can provide a satisfactory solution in the absence of any theory of the problem. ), much much before I had them :).
Recently, at Microsoft, I built a tool for analysing telemetry data, originating from web applications and services. The tool models several practical problems of interest (e.g. Identifying top workflows, Behavior based anomaly detection, Identifying top deployment slots, etc.) in terms of machine learning techniques such as Topic Modeling, Time Series Decomposition, Streaming Least Squares etc. At Microsoft, I also won four Hackathons (at different levels: Global, India, Org & Campus Hires) with different teams, for building ML powered software systems. I also developed an algorithm that reduces word embedding dimensionality by 50% while preserving its performance. The work was accepted for poster presentation at NIPS LLLD Workshop.