I created a semi-supervised classifier for time-series photon measurements of supernova called light curves. The neural network consisted of a variational autoencoder and a random forest classifier. The autoencoder was first trained and then the random forest would classify light curves into 8 different classes based on the light curves encoded representation. I first processed all the data by using min-max normalization and added padding to make all light curves the same size. Because of this padding, I then added masking so that the autoencoder skips the time steps with not data. Lastly, I tuned all the hyperparameters of the neural network. I was able to reduce the light curves to 30 data points and get a 93% accuracy rate.

The model is supposed to mimic collecting several pieces of data near the Azores Triple Junction using a command ship, aerial drones (UAVs), and autonomous submarines (AUVs). Improved the logic of the model by changing the implementation of the UAVs and AUVs to use two arrays instead of six and fixed how agents were chosen to go and collect data by optimizing the algorithm’s path selection for agents Added all missing documentation to the project and created all unit and expect tests to ensure the model can handle extreme cases before deployment using JUnit. Published on July 29th, 2022

Learned OCaml by first creating a simple game of Snake. I then extended it significantly by adding a leaderboard, home screen, multiplayer version and then made it into a client-server distributed system.Created a trading bot that partook in an electronic trading challenge (ETC) where bonds, ETFs, and American Depository Receipts (ADRs) were traded. Strategies I implemented were hedging and fading based on fair values that my bot estimated, and won 2nd place in the competition between the other fellows. Extended the Quickcheck library, which allows one to create quick unit-tests. The library has generators, which return random data to write tests; shrinkers, which when a test fails, finds a smaller counterexample; and observers, which are generators but for higher-order functions. I implemented all three for functional queues and deques, lazy strings, and error messages.

Planned and installed irrigation systems in the greater Houston area, specifically, the Woodlands; Conroe, TX; River Oaks and Memorial City

- Aided in creating a new bookkeeping system using Quickbooks and improved customer retention by creating two separate pricing systems: one for landscaping companies and another for the public. - Reduced water waste by creating a more efficient watering schedule by using Python functions that optimized water usage determined by weather forecast, type of shrub and dewpoint