STATEMENT OF PURPOSE Almost 26 years ago, on the midnight of Dec. 10, 1993, an executive at id Software uploaded a 2.39 MB file to the University of Wisconsin’s FTP server. This small, insignificant moment in history went on to completely revolutionize modern gaming. The name of the file was doom1_0.zip, the first copy of the classic first-person shooter game, Doom. Following the years since its launch, Doom eventually made its way around the world, impacting the lives of several players, including me. Before I could even barely feed myself with a spoon, I started playing Doom, which sparked my lifelong passion for video games, computers and eventually, artificial intelligence. Naturally, I chose to pursue my Bachelors in Computer Science Engineering at SRM Institute of Science and Technology, Chennai. It was during this time I had a peek into the field of artificial intelligence and machine learning. While artificial intelligence present in video games hooked me into the field, it was the fact that we could “train” computers to automate particular tasks, that reeled me in. I was clearly interested in teaching a computer how to play video games. Akin to almost every other enthusiast in this field, I started my journey into this field by enrolling in Prof. Andrew Ng’s online Machine learning courses on Coursera. These courses gave me the necessary foundation for me to get working on my pet projects right away. Deciding to put my skills to use, I participated in the Tata Crucible Hackathon, a national level hackathon organized by the Tata group. One of the challenges was to count the number of straws in a bundle from only pictures, for which my team and I built a very basic object detector using TensorFlow. Unfortunately, we were unable to win the hackathon but the host liked our solution very much and invited us to work on their actual problem; counting rebars for their steel warehouses. This marked the beginning of a year-long internship with Tata where the majority of my work experience in machine learning, particularly computer vision was gathered. Donning safety jackets and taking pictures of rebars at a Tata steel warehouse gave me very valuable experience in collecting and preprocessing data. In the end, we delivered a working model that could count rebars of various sizes with almost 93% accuracy. It was also during this time, I had the opportunity to attend a 21-day training program on “Big Data Analytics using Artificial Neural Networks” at the National University of Singapore, in collaboration with Hewlett-Packard. Partaking in this event has been one of the most rewarding experiences of my life, as I had the opportunity to interact with the professors over there and catch a glimpse of the research they were doing. It gave me a good idea about the problems being tackled in the international area. Right after my internship at Tata, I joined Nyx Wolves - a business solutions startup, as their lead machine learning developer. I was involved in developing a computer vision solution to count the number of vehicles on Indian roads and automatically detect and tag traffic violators for the Chennai Traffic Police. I developed a way to individually count vehicles over multiple classes (car, bus, etc.) across different lanes. It was a few months ago, during an intense gaming session, I realized that games have more key bindings that we have fingers and wondered if it was possible to ‘map’ certain keys to voice commands. I started working on Shadowcol - a control scheme using siamese networks to play video games with voice. I decided this would be my capstone project for my undergraduate study and have been working on it under the guidance of Dr. C.N. Subhalalitha. Although all the projects I’ve been a part of definitely make use of machine learning and in a broad sense, artificial intelligence, I realized I was doing nothing related to what actually seduced me into this field; making my computer play video games. With that in mind, I’ve recently begun looking more into the field of reinforcement learning. Brilliant research is already being done in this field, with OpenAI’s bots defeating the world champions in Dota 2 and Deepmind’s AlphaGo defeating Lee Sedol in Go. I’m specifically curious as to whether we can “teach” a computer to play more complex, story-driven games. So, my current goal and pet project is to build an AI agent to play the latest iteration of the game that drove me to this field, Doom. While I understand that such a project will not directly aid society in any reasonable way, I’m confident that the implications of such a project will extend beyond just video games. Beneath its surface of mindless violence, Doom taught me a lot of things in life, particularly that of perseverance. It has not been an easy journey so far since most of it involved consuming an unhealthy amount of caffeine and yelling at my computer when my code doesn’t work. Therefore, I do not expect my graduate studies at {INSERT COLLEGE NAME} to be easy either. The university has {SPICE ABOUT UNIVS ACHIEVEMENTS IN ML AND STUFF}. I’m particularly excited about {SOME COURSE OR SOME PROF YOU LIKE}. I’m confident {INSERT COLLEGE NAME} will equip me with the right knowledge and skillset to realize and achieve my goals.
Enter the password to open this PDF file:
-
-
-
-
-
-
-
-
-
-
-
-