We developed a computer vision pipeline using Python, OpenCV, and DeepFace to enable real-time facial expression analysis in the social robot Epi. We validated the system through a controlled experiment comparing algorithmic outputs with human observers, identifying both hardware constraints and software limitations in the aforementioned HRI technology.
Read full descriptionA custom, real-time neurofeedback software was built from the ground up. Designed for the Muse 2 EEG headband, the software analyzed alpha/theta ratios, translating biological metrics into real-time feedback cues. This system was used to benchmark embodied vs. audio guidance, ultimately testing whether a robot's physical presence enhances or hinders meditative states.
Read full descriptionDesigned a machine learning system to predict UFC outcomes by quantifying complex variables like fighter durability and recovery rates. Moving beyond basic physical attributes, the algorithm leverages these inputs (including quantifiable features I engineered) to find high-value bets where the market undervalues specific matchups. The model has demonstrated significant predictive power, achieving a 71% ROI (on underdogs) in backtesting.
Read full descriptionA collaborative philosophical inquiry on the ethical consequences of integrating AI into human cognitive life. Drawing on frameworks like the Free-Energy Principle, we argue that widespread reliance on AI for 'augmentation' rather than 'restoration' risks inducing cognitive atrophy, a gradual erosion of essential human capacities such as autonomy, competence, and authentic connection.
Read full descriptionA custom-coded portfolio built without external libraries using solely HTML/CSS. The website was built to be a distraction-free environment for long-form documentation.
Read full description