Nathan Babcock

Implemented the React front-end for a "play for purpose" NFT game called Zeedz. The game UI implementation includes user login and signup, purchase and unboxing experience, and inventory management for the player's digital collectibles stored on the Flow blockchain.

The front end connected to the team's existing back end, including a custom GraphQL API, AWS Cognito, integration with the Flow blockchain via the Flow Client Library, distributed file hosting with Pinata, and a few other services.

Teamwork was fully remote over Google Hangouts, mostly with Germany and the UK collaborators.

A machine vision AI application that automatically detects gaming highlights in live or recorded video and records 10 to 30-second clips. I built the core engine in Rust. It decodes video with FFMPEG on the GPU. It then uses one of two computer vision methods, either custom-trained TensorFlow models or a more general-purpose OCR (optical character recognition) using an open source Google library called Tesseract. The pipeline runs at about ten times the speed depending on the GPU. It can process 10 seconds of the input video in about one second.

In addition to the core engine, I built a native Windows front-end in a relatively new framework called Tauri, similar to Electron, but integrates with Rust and uses the system WebView instead of bundling an instance of Chrome. The front-end is also built using Solid.js, a framework similar to React but focused more on high-performance and well-suited for this performance-heavy video processing and machine learning application.

Finally, I built a website to distribute and market this tool. It can be viewed at Hypetrigger.io. I built it using Next.js, and it is 100% statically generated. The content is created from JSON configuration files and a third-party REST API.

A first-person 3D maze exploration game written entirely in JavaScript, React, and WebGL, using Three.js and react-three-fiber. The game features lighting and shadows, collision detection, and post-processing effects like bloom and anti-aliasing. The prototype is free to play on itch.io.

Using Google Drive as a CMS leverages Google Drive's public REST API to source content from the content-serving back-end of a dynamic website. A companion article for this library was featured in a CSS Tricks article.

Lux AI is an AI programming competition hosted on Kaggle. I built a deep neural network that used imitation learning to mimic replays of the best public bots. This neural network achieved 80% accuracy in classifying optimal game moves. As a result, the entry ranked in the top 15% of the leaderboard.

This university project involved terrain elevation data for Lake Tahoe, California, converted into 3D geometry in Unity for a virtual reality project. The VR environment was a "CAVE" (Cave Automatic Virtual Environment), which consisted of three walls and a floor, with projectors displaying the rendered environment onto each surface. The viewer could fly around this virtual landscape using a game controller.

A 2D, physics-based idle game in the incremental game genre. I implemented the 2D collision physics and rendering in PixiJS. Gem Drop won third place in a Kongregate monthly contest and has been played more than 75,000 times.

A multiplayer territory control strategy game written using WebSockets and Canvas. The website accepted payment using Stripe API and users logged in with Auth0. The website server was hosted on AWS with Nginx and Docker.

This tool pulls GeoJSON data for any location from Open Street Map. It uses the data to simulate the radio coverage for various deployments of LTE or broadband radio antennas, showing blind spots and dead zones and helping determine optimally efficient configurations of hotspots.

Developed internally for Motorola Solutions – the specific algorithm and source code are proprietary.

A project exploring various technical indicators on historical stock market datasets. The stock data was pulled from Yahoo finance, and the indicators included different combinations of moving average, MACD, RSI, and others.