Hi, I'm Zach Hoffman

I designed the model in Fusion 360 and 3D printed it. It has 4 DOF, and it is purposed to hold high torque servo motors and a raspberry pi camera.

‹ ›

Each square is classified using a custom CNN built in PyTorch to determine piece color (white / black) or empty. I then compare consecutive board states to detect the opponent's move.

• CNN architecture: input 64x64 square images → 3-class output (white / black / none)
• Board state is tracked using python-chess, updated after each move detection
• One limitation is that the current implementation assumes the game begins in standard starting position, which means it can't yet handle arbitrary mid-game setups

Integrated the 4-DOF arm with high-torque metal-gear servos (40 kg-cm - 60 kg-cm) via a PCA9685 PWM driver and Raspberry Pi.

• Hardware build complete; gearbox & servo mounts designed for strength and repeatability
• Sadly, this is still a work-in-progress. I have basic servo control working, but the inverse kinematics and pick-and-place routines are not yet finished.

Players can freely draw race tracks using a Pygame interface, enabling completely unique circuits to be built.

• Mouse-based track drawing with real-time rendering.
• Brush sizes 1-5 and eraser tool for fine-tuning.
• Track boundaries are converted into collision masks for the simulation.

The simulation uses a simplified yet realistic car physics model that incorporates traction, downforce, and mass to influence velocity and steering angle. Each car calculates maximum turning capability dynamically at runtime.

• Physics include traction loss, momentum, and cornering limits.
• Throttle position directly affects acceleration (not time-based).
• Downforce and mass control grip and turning radius at higher speeds.

Over time, cars learn to brake earlier, accelerate out of corners, and take smooth apexes. Observing the evolutionary process reveals natural emergent driving behaviors similar to real racing strategies.

The site is a full-stack wedding platform built entirely on the Cloudflare ecosystem, combining Pages, Workers, D1, R2, and Resend for a serverless, globally distributed experience.

• Frontend: React + TypeScript + Tailwind (Cloudflare Pages).
• Backend: Workers with REST API routes.
• Storage: D1 for relational RSVP data and R2 for photo uploads.

A collaborative photo-sharing space where guests can upload and browse images from engagement events, the ceremony, or the reception; all stored securely using Cloudflare R2.

• Guests are able to upload photos directly through the website.
• Uploads are verified and stored in Cloudflare R2.
• Uploaded photos are held private until admin approval (toggleable), and tunable rate limiters are in place.
• Admin notifications sent automatically via Resend when new uploads arrive.

A private admin portal gated by Cloudflare Access allows managing guest lists, checking RSVP stats, and sending announcements.

• Dynamic charts showing RSVP rates over time.
• Editable tables for manual corrections.
• Integrated email-sending from the dashboard interface.

A 2D wave-style shooter where a pawn defends against spawning chess-piece enemies. Players use collected board labels (letters and numbers) as weapons.

• Designed as an arcade-style infinite survival game.
• Combines action mechanics with chess-inspired visuals.
• Built entirely with Pygame using object-oriented design.

The project follows an event-driven structure with distinct classes handling logic, input, and rendering. Each subsystem is isolated for clarity and modularity.

• GameLoop: handles frame updates, collisions, and wave progression.
• MainMenu & ControlsMenu: user interface and navigation.
• Sprites & Textures: manage all graphical assets and animations.
• UsefulFunctions: utility layer for input, scoring, and collision checks.