Resume
GitHub
LinkedIn
Resume
GitHub
LinkedIn
A command-line–driven, multi-user blog aggregator backend that collects and organizes posts from RSS feeds in real time. Users can register, log in, add and follow feeds, and browse aggregated posts through a structured set of commands, while a background process continuously scrapes and updates content. The system is designed with a modular command architecture, authenticated workflows, and a relational data model to ensure data integrity and scalability. Feed data and posts are automatically deduplicated and kept up to date, allowing users to consume content from multiple sources in one place without manual tracking.
A minimal, interactive Tarot reading web application that allows users to ask questions, draw Tarot cards, and receive AI-generated interpretations in real time. Each reading is dynamically personalized by randomly selecting cards from a full Tarot deck and embedding them into the prompt, creating a unique experience on every interaction. The application emphasizes performance and user experience through server-to-client streaming. AI responses are streamed incrementally using HTTP event streams and rendered with a custom typewriter effect, allowing text to appear as it is generated rather than waiting for full completion. This approach significantly reduces perceived latency and makes interactions feel responsive and engaging. Built with a modern React and Next.js architecture, the project cleanly separates UI components from business logic and reusable hooks. It demonstrates practical use of streaming APIs, prompt construction, and client-side state management, while maintaining a lightweight, focused design that highlights real-time interactivity over unnecessary complexity.
An interactive Pokédex application that combines real-time 3D graphics with a stateful backend to simulate Pokémon capture mechanics. Users select Pokémon, interact with a fully animated 3D Pokéball, and attempt captures that follow probability rules based on each Pokémon’s rarity. The frontend focuses on precise state management and visual feedback, using a finite phase model to coordinate animations, user input, and reveal timing without invalid states. Subtle animation details, such as shake intensity and reveal sequencing, are tightly coupled to game logic to create a responsive, game-like experience in the browser. The backend exposes custom endpoints to manage capture logic, inspection, and persistence using an in-memory state model. Pokémon data is fetched from the PokéAPI and cached after successful captures to avoid redundant requests. The system is containerized and deployed with a production-ready setup, demonstrating end-to-end integration between interactive 3D UI, backend game logic, and cloud infrastructure.
A modern, multiplayer take on the classic Asteroids arcade game that separates game simulation from rendering. The core game loop runs as a headless Pygame simulation on the server, while real-time game state is streamed to a browser-based client over WebSockets for responsive, low-latency gameplay. The backend manages player connections, input state, physics, collisions, and game progression within a shared simulation loop, allowing multiple players to interact in the same game world simultaneously. By running Pygame in a headless environment, the system remains compatible with cloud infrastructure while preserving deterministic game logic on the server. The frontend renders the streamed game state using HTML5 Canvas and a requestAnimationFrame loop, ensuring smooth animation synchronized with the browser’s refresh rate. Deployed with a containerized backend on AWS EC2 and a static frontend on Vercel, the project demonstrates real-time networking, server-authoritative game design, and production-ready deployment practices.
A lightweight, backend static site generator written entirely in Python that converts Markdown files into complete HTML pages without relying on third-party libraries. The project implements a custom Markdown parser and an HTML node tree to transform structured text into valid, reusable HTML components. The generator recursively walks a content directory, applies a shared HTML template, and preserves the original folder structure when producing output files. This approach enables scalable site generation from simple Markdown content while keeping build logic explicit and easy to reason about. Designed with correctness and maintainability in mind, the project includes comprehensive unit tests covering Markdown parsing and HTML rendering logic. It demonstrates core backend concepts such as parsing, tree-based rendering, recursive file processing, and test-driven development in a clean, dependency-free Python codebase.