My Projects

Selected projects showcasing my expertise and capabilities

CodeArena

Active

Project Overview

CodeArena is a full-stack competitive coding platform in the style of LeetCode. Candidates browse problems, write solutions in a Monaco-powered editor, submit, and receive real-time verdicts over WebSockets. An Angular admin panel lets operators author problems, manage test cases, and review submission logs — all backed by the same Go API.

Key Features

  • Monaco editor with C++ and Python syntax highlighting
  • Real-time submission results via WebSockets (no polling)
  • JWT authentication — register, login, submission history
  • Sandboxed C++ judge using fork/exec, setrlimit, and SIGKILL timeouts
  • Angular admin panel — problem CRUD, live Markdown preview, test case manager
  • 61-test Go suite — handler tests, store unit tests, mock judge

Technical Details

Candidate UI

React + Zustand + React Query

Admin Panel

Angular 21 (standalone, signals)

API Server

Go (stdlib net/http, pgx, gorilla/websocket)

Judge

C++17 (fork/exec, rlimit, nlohmann/json)

Database

PostgreSQL

Auth

JWT HS256 (golang-jwt/jwt)

Technology Stack

React Angular Go C++17 PostgreSQL WebSockets JWT Docker Fly.io
Live Demo View on GitHub

AgentArena

Active

Project Overview

AgentArena is a reproducible research lab that pits Claude model variants against each other and against classic rule-based strategies in iterated game-theory tournaments. Models play hundreds of rounds of games like the Prisoner's Dilemma, and every move, reasoning trace, token count, and dollar cost is logged for analysis. The findings ship as a series of data-driven LinkedIn articles.

Key Features

  • Round-robin tournaments across Sonnet and Haiku variants plus 7 classic bots
  • Orthogonal prompt personas (neutral, cooperative, selfish) and memory windows
  • Analysis pipeline: score tables, cooperation-over-time, forgiveness and retaliation charts
  • Cost guardrails with prompt caching and a pre-run spend projector
  • Key finding: the prompt swing outweighs the model swing in cooperative play
  • Fully reproducible runs, with article footnotes pinned to commit and run IDs

Technical Details

Language

Python 3.12+

LLM SDK

anthropic (prompt caching, Batch API)

Analysis

pandas + matplotlib / plotly

Tooling

pydantic v2, typer, pytest, ruff, mypy --strict

Technology Stack

Python Anthropic API pydantic pandas matplotlib Typer pytest Game Theory
Read the Article View on GitHub

Materialist

Active

Project Overview

Materialist is a comprehensive Flutter application designed to streamline material verification for subcontractors. The system enables teams to track material deliveries in real-time, manage project inventories, coordinate team members, and maintain accurate records through an intuitive cross-platform interface.

Key Features

  • Real-time material tracking and verification
  • Team member management and coordination
  • Project-based material organization
  • Advanced search and filtering capabilities
  • Task management system
  • Multi-language support (EN, HI, TA, UR)
  • Cross-platform (Web, Android, iOS)
  • Real-time database synchronization

Technical Details

Framework

Flutter 3.41.9

Language

Dart 3.11.5

Backend

Firebase Firestore

Authentication

Google Sign-In

Technology Stack

Flutter Dart Firebase Firestore Go Router Google Sign-In Flutter Animate PWA

Platforms

Web
Android
iOS
Try Demo View on GitHub Local Repository

Sync Bot

Completed

Project Overview

Sync Bot is a Python-based Discord bot built for the Sam Houston State University Dungeons and Dragons Club. It bridges Discord and SHSU's OrgLink platform to automatically sync member roles and surface event activity, eliminating the manual role-management overhead the club's officers were previously stuck with.

Key Features

  • Automated Discord role sync against OrgLink membership
  • Reduced manual administrative work by 90%
  • Automated event notifications and reminders
  • Increased member engagement by 40%
  • Boosted event attendance by 25%
  • Built on the discord.py async API client

Technical Details

Language

Python 3

Library

discord.py

Integration

OrgLink (SHSU)

Timeline

Sep 2022 – Dec 2023

Technology Stack

Python discord.py asyncio REST APIs Automation

Platforms

Discord
View on GitHub

EFCoreLint

Active

Project Overview

EFCoreLint is a Roslyn diagnostic analyzer that catches common Entity Framework Core anti-patterns at compile time. It ships as a NuGet package and surfaces squiggles directly in the IDE — no runtime overhead, no configuration required.

Diagnostics

  • EFLINT001 — Client-side filtering after materialization (Warning + code fix)
  • EFLINT003 — Read-only query missing AsNoTracking (Info + code fix)
  • EFLINT005 — EF Core async method called without await (Warning + code fix)
  • 22/22 unit tests passing
  • CI/CD pipeline: builds and tests on every push, publishes to NuGet on release tags
  • Published to NuGet as v0.1.0

Technical Details

Language

C# / .NET

API

Roslyn (Microsoft.CodeAnalysis)

Testing

xUnit + Roslyn analyzer testing

Distribution

NuGet (v0.1.0)

Technology Stack

C# .NET Roslyn Entity Framework Core xUnit NuGet GitHub Actions
View on GitHub View on NuGet

enter-giveaway-inator

Case Study

Project Overview

enter-giveaway-inator discovers Pokemon card giveaways on TikTok (and Whatnot), parses the entry requirements out of each post, and notifies you so you enter by hand. It is equally a case study in engineering judgment: built, tested, and deployed to the cloud running 24/7, then deliberately wound down once the core goal of real-time live giveaways proved structurally unsolvable on a hobby budget. The lesson turned out to be the deliverable.

Key Features

  • Scans TikTok and Whatnot, parses follow / like / comment / tag / deadline rules
  • Notifies via console, desktop, Discord, and phone push (ntfy), live entries first
  • Pluggable sources and notifiers: a new platform or channel is one drop-in file
  • Scoped to discovery, not auto-entry, by design (ToS-safe, keeps entries valid)
  • Diagnosed and worked around IP blocks, container OOMs, and pay-per-result costs
  • 39 offline tests; the suite never touches the network

Technical Details

Language

Python 3.11+

Scraping

Playwright + Apify actors

Deployment

Fly.io + Docker

Storage

SQLite de-dupe on a persistent volume

Technology Stack

Python Playwright Apify Fly.io Docker SQLite Discord ntfy
View on GitHub

Coming Soon

I'm always building. New projects will appear here as they launch.