Projects

babygrad

A small neural-network library built from first principles in pure Python. It progresses from tensors and automatic differentiation through training, residual networks, and transformers.

Webcam Eye Tracker

An end-to-end deep-learning method to predict real-time eye gaze locations. Unlike commercial eye trackers that rely on infrared, this method uses only a regular webcam.

Warden

Modular security boundary for integrating containerized AI agents into 3rd party applications. Use as a standalone agent orchestrator, or as a sidecar HTTP API to control the agent environment.

Agent-ready Portfolio with WebMCP

Working WebMCP example built into this portfolio. It gives visiting agents structured tools for interview contact, project research, and evidence-based role-fit questions.

Game Review Aggregation

ETL pipeline to collect game reviews (web scraping, APIs), and transform / store them for analysis. Fed into a machine learning pipeline for clustering and sentiment analysis.

Parcel

A companion app for your AI agent that keeps your job search organized. Track jobs, postings, resumes, notes, and interviews in one workspace, with reusable career context your agent can use across applications.

sensormotion

Open-source Python package to aid the analysis of sensor-based motion data. Includes algorithms for peak detection and extraction of various gait metrics.

Cognitive Battery

Developed a modular data collection platform to gather experiment data for various Cognitive Science computer tasks. This application was created to save time during my PhD.

RoK Talents

Developed a web application that allows Rise of Kingdoms players to design and customize character talent builds. Builds can also be linked and shared with other players.

Color Blind Simulator

Internal research tool for simulating color blindness in video games. Real-time simulation allows for quick identification of accessibility issues in games research.

Smartphone Gait Measurement

Developed novel smartphone-based method to measure outdoor gait kinematics. Used multilevel modelling to analyze multidimensional sensor and image data.

Loot Box Reward Distribution

Used statistics and hypothesis testing to answer a common question in the gaming community: do different methods of opening loot boxes lead to a different distribution of rewards?

HoloLens: Learning in Mixed Reality

Designed experiments, collected data, and performed statistics to understand how users learn brain anatomy and move around in mixed reality environments.

Dual Eye Tracking

Developed technique to perform live eye tracking on 2 subjects simultaneously. Analysis consisted of various auto-correlation and cross-correlation methods.

RoK Commander Pairings

Created force-directed graphs to illustrate ideal pairings of Rise of Kingdoms game characters. The visualization provides insight into how to optimize team compositions.

Overwatch Performance

Analyzed and visualized trends in personal Overwatch skill rating over time. Breakdowns by game mode and map type help to identify factors that affect performance.