Software Projects
The following is a listing of notable software projects that I have worked on.
Advent of Code 2023
This one was just for fun! My branch at NASA JSC likes to take place in a friendly competition for Advent of Code (AoC) - an annual coding challenge consisting of 25 days of problems with increasing difficulty. AoC is intended to be a fun learning experience for all, and indeed I myself learned a lot about data structures, algorithmic efficiency, and more. I even took home first place on our local leaderboard! This came with the added perk of some friendly banter for proving Python to be superior to the other languages in the mix - Fortran, C++, Julia, and even Excel/VBA.
Neural Network Predicting COVID-19 Impact on Electric Grid
This was a final project for my graduate Machine Learning course while at Stanford. Using the Keras package, we trained a Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) to predict the demand on the local electric grid in a number of major American cities. By training our model on pre-COVID load data, we predicted that we would be able to observe the effect of COVID-19 as the difference in predicted vs true load magnitude. Indeed, we observed a clear delta in predicted vs true electric grid demands that coincided with the onset of COVID-19 shutdowns.
A Reinforcement Learning Algorithm for Blackjack
This was another final project at Stanford, this time for a course entitled Decision Making Under Uncertainty. For this project, we applied a variety of different reinforcement learning algorithms to learn optimal policies for playing the popular casino card game - Blackjack. We demonstrated that the resulting policies were far superior to random policies and were nearly identical to “truth” values (i.e., reference Blackjack strategy charts for optimal gameplay).
Fuel-Optimal Orbital Transfers to Mars
This was an open-ended final project for my graduate school course Engineering Design Optimization. For this project, I chose to apply a variety of different optimization algorithms to the problem of finding fuel-optimal orbital transfers to Mars. By making use of Lambert’s Algorithm, I was able to model the dynamics of the problem across a range of launch/landing dates. I also successfully validated my results by showing that they aligned closely with the actual launch/landing dates of the recent Mars 2020 mission. My favorite optimization algorithm I used was the Cross-entropy algorithm - a stochastic population-based algorithm - due to its simple and elegant use of probability functions!
