Here are more detailed descriptions of the projects I have been involved in so far, from most to least recent.
Semantic Based Prediction Planning - Parasol
My most recent and newest project. For effective collaboration, robots must do more than just react; they need to anticipate. This requires inferring unobservable human intentions and predicting future actions. Anticipatory behavior is the next frontier in fluent human-robot collaboration (HRC), moving from reactive to proactive robots. This project aims to synthesize current prediction methods with core motion planning methods, as current prediction methods such as Bayesian methods, Markov Models, and Supervised Deep Learning are no longer sufficient. We aim to solve those utilizing planning methods.
HAS-RRT: Motion Planning for Protein-Ligand Interactions - Parasol
Last semester, my project was aimed at using Hierarchical Annotated Skeleton RRT (HASRRT) to evaluate whether binding sites in proteins are accessible. A key principle of drug design is how ligands bind to proteins, and how those interactions determine the behavior of the protein. We planned to utilize HAS-RRT to help ligands traverse the configuration space within a protein to find the lowest-energy path to their target binding site. Compared to other motion planning algorithms (Basic RRT, DR-RRT, EET, etc.), HAS-RRT has shown significant runtime and efficiency improvements.
Drive Doctor
Working closely with a team of 3 to develop a Google Chrome extension that manages, organizes, and creates various Google Drive files utilizing Python, HTML/CSS, JSON, and JavaScript. We utilize Flask to connect the front end and back end of the project, adding the functionality of creating new files for the extension. We also created an LLM utilizing the mini-LM Sentence transformer from Hugging Face to compare the semantic similarity of files to effectively organize them into folders
Miscelaneous Machine Projects
Within the multiple classes I have taken at UIUC, including CS128, CS225, and CS233, I have had the opportunity to complete dozens of machine projects. These helped me not only solidify my understanding of data structures and algorithm design, but apply them in a practical sense. My top three so far are Seam Carving (resize images without loss of quality utilizing the seam carving method), Mazes (builds random mazes utilizing disjoint sets), and Puzzles (animated a solution of a randomly generated 15 puzzle). All of these, and more, can be found on my personal GitHub.