Hi, I'm Cameron

I build scalable platforms, integrate AI, and turn abstract ideas into reality

I care about cohesiveness, accessibility, and impact in my work

See selected work Contact

Artificial Intelligence

Systems Thinking

Fullstack Development

Data Science

Research

My Projects

I've been working on a range of projects over the years, from playful ideas to AI-powered tools.

Game development first taught me to enjoy the process of building. Now, after spending more time with AI and product work, I'm even more interested in how software can create real impact.

GitHub

Panda Non-Prehensile Sorter

Robotics
Reinforcement Learning
MuJoCo
JAX
C++
Three.js

Overview

A robotic arm trained with reinforcement learning to push moving packages from a conveyor into color-coded bins without grasping them.

My contribution

I built the MuJoCo simulation and PPO training pipeline with JAX, including curriculum learning, reward shaping, and parallel physics workers with C++ acceleration. I also created an interactive browser demo that runs the trained policy and physics locally with Three.js rendering.

Try demo LinkedIn post

Bone Fracture Malunion Research

Data Science
Machine Learning
Data Visualization
Research
Pandas

Overview

A research project analyzing the MIMIC-IV medical dataset to predict bone fracture malunion outcomes using machine learning techniques.

My contribution

I used Pandas and JMP Pro to clean and analyze the data and built predictive models using PyCaret. I also first-authored an abstract that was accepted for presentation in the ASBMR and the BMES 2025 conferences.

See abstract View code

PatchSight

AI Tooling
Playwright
Vision Models
Frontend
Human-in-the-loop

Overview

An experimental AI-powered visual debugging system that uses Playwright screenshots and multimodal vision models to identify layout issues in rendered web interfaces.

My contribution

I built the MVP workflow for screenshot capture, structured visual feedback, iterative patch generation, and a human-in-the-loop dashboard for reviewing scores, feedback cards, iteration history, and applied file diffs before changes are committed.

View code LinkedIn post

MedSnap

Next.js
MariaDB
OpenAI Integration
LLM
Data Processing
MUI

Overview

An AI-powered medical report autocompletion tool made for streamlining repetitive notetaking.

My contribution

As a fullstack developer, I handled API routing, user authentication, database security, and UI efficiency. I implemented the autocompletion feature using OpenAI's API.

Visit site

Drone Recovery Reinforcement Learning

Reinforcement Learning
Autonomous Systems
StableBaselines3
Research

Overview

Quadcopter using reinforcement learning (PPO) in Unity environment to recover after a single-rotor failure.

My contribution

I integrated Microsoft's AirSim with Unity and StableBaselines3, using Tensorboard for visualization. I worked on reward shaping and hyperparameter tuning to achieve stable learning and successful recovery behaviors.

View code