I have worked on several projects apart from my doctoral research. These projects span a wide domain of my interests, from sustainable design, to optimization, to web development. For some of the projects, details are available on dedicated pages, explaining the project in more detail, delineating some of the challenges, and describing the solutions developed. The projects are presented below (in no particular order).
Solar Decathlon Europe 2014 was a competition organized by CSTB France to design, construct and operate a life-size zero energy solar powered house. I cofounded Team Shunya along with 6 other students at Indian Instiute of Technology Bombay (India), the first team to be selected to participate in SDE 2014. My role in this project was as a construction manager and site operations coordinator, other than being involved in the design conceptualization, designing process and team management. As the construction manager, I developed a schedule for assembling the house at the competition venue within 10 days. As the site operations coordinator, I led the night shift, comprised of 12 students, during the assembly of the house at the competition. The team received a special mention for sustainability and placed 2nd among the first time participants. This project developed the skills required to aproach a never before attempted large scale project and devise and execute solutions to real-time problems under pressure.
The aim of this project was to develop a schedule for the NFL based on a measure of the popularity of each match with the overall objective being to maximize the total popularity of the schedule. The optimization was an integer optimization problem with over ten thousand variables which was solved using the branch and bound algorithm implemented via Gurobi and Python. In order to optimize, I had to first develop an objective function, which was basically a methodology to estimate popularity of each match. Once this was developed, about 20 families of soft and hard constraints were imposed. This project helped develop the skills required for approaching an optimization problem from start to end.
In this project, the cross-section of a beam (a structural member which resists bending forces in buildings) was optimized with weight minimization as the objective function. The problem was modeled as a parameter optimization problem with each parameter corresponding to a dimension of the beam. A program for EA, previously developed by Dr. Kenneth DeJong, was modified to solve this optimization problem. Several hard and soft constraints were implemented to keep the dimensions of the beam within reasonable bounds. The final result obtained verified the design solutions used in practice.
Currently developing a python based program to predict the next word in a sentence. The algorithm currently being implemented is very simple and uses the words from a sample set of text files (called corpus, corpora in plural) to determine a probabilistic distribution of the words given the previous two words (can be expanded to more than the previosu two words, but becomes computationally challenging). The problem can effectively be modeled as a markov chain in which the probability of the next word is dependent on that of the previous word. The probabilistic distribution determined is then used to suggest the next word. More details to this project would be added once completed.
I worked as a web developer and developed, along with another intern, a hedge fund management web application for the client. MVC 3.0 architecture, implemented via the .NET framework, was used to keep the hedge fund data, the program logic and the UI independent of each other. An algorithm was developed to parse mathematical expressions from the database. The expressions, once parsed, yielded formulae to compute the value of desired parameters for the hedge fund.