Portfolio

Projects, research, games, and experiments.

What ties it together

The tools and topics change, but the pattern is usually the same: find a real point of friction, make the tradeoffs visible, and build something useful enough to test in the open.

Browse the portfolio

More builds and experiments live on GitHub.

Favorites

A few projects I come back to first.

These are the clearest examples of how I like to build: real constraints, technical depth, and enough care in presentation that other people can follow what is going on.

Apps/Tools

A planning tool for UMD students that brings course search, degree requirements, and four-year planning into one place.

Demo mode is handled at the data layer, not with per-component conditionals
Degree audit maps completed courses to requirements in real time
CI runs type checking and linting on every pull request

ML/AI

A Blokus Duo self-play research project using PPO, MCTS, and CNN-based policy and value networks.

Self-play PPO with MCTS-guided data collection on a 14×14 board
Fixed reference opponents keep evaluation honest
MCTS simulation count changes over training instead of staying fixed

Other projects, demos, and experiments.

This is the broader mix: class projects, research builds, tools, and smaller experiments that still taught me something worth keeping.

Games

A cross-platform maze game built around procedural generation, daily puzzles, and replayability instead of fixed handcrafted levels.

Recursive backtracker generator with flood-fill solvability checks
One C# codebase runs on iOS, Android, and Windows through .NET MAUI

Apps/Tools

An iOS AR scavenger hunt built for a live university event, where first-time users had to pick it up and move immediately.

Persistent HUD keeps progress visible during real-world interruptions
SwiftData keeps progress alive across app restarts

Robotics/Hardware

Soft robotics research covering the printed SpiRob body, Arduino cable control, and simulation environments for a logarithmic-spiral grasping robot.

relax_ratio = 0.2 was tuned experimentally to avoid over-tension
Bringup sequence moved from single motor tests to coordinated control

Technology landscape

What I used to build them.

Hover a project to see the tools around it, then click through to the case study. You can also scroll to zoom and drag to pan.

Apps/Tools Games ML/AI Robotics/Hardware Technology

Projects, research, games, and experiments.

Browse the portfolio

A few projects I come back to first.

Orbit UMD

BlokusDuo ML Agents

Other projects, demos, and experiments.

Maze Escape: Pathbound

Henson Day AR App

SpiRobControls (Drexel Zheng Lab)

What I used to build them.