Hi, I'm Joe Gibbs.

I'm a software engineer. Check out some of my projects below.

Contact me

Career

Yaarn

2022 —

Starting in 2022 I have been working on Yaarn, an AI-powered public relations management platform.

SEN

2021 — 2022

From 2021 I worked at SEN, working on internal software such as scheduling systems.

I cofounded Grouptag in 2020. My goal with it was to improve the way that Gen Zs can network together through an online platform that connects people together based on their skills, goals and interests.

Treiner

2019 — 2020

I joined Treiner in 2019 as CTO. Treiner is a platform that helps soccer players find affordable coaches near them, and helps coaches find new players through avenues that they might have otherwise overlooked.

Projects

Channel 784

2023

Channel 784 is an AI-generated news parody. Using Unreal Engine and GPT3, it generates scripts for a fake news program, voices and animates them, with debates, weather and graphics for each segment. You can view some of its episodes here.

Fall of an Empire

2020 —

Fall of an Empire is an Unreal Engine strategy game. The player must balance the needs of their empire's provinces while also fighting off barbarians who want to take your cities. You can read more about it here.

Redland

2020

Redland is a procedurally-generated, browser-based HTML Canvas game about gathering resources, building equipment and discovering pieces of a map to find treasure. I developed it with my colleague Samuel and you can play it here.

Forlorn Desert

2019

Forlorn Desert is a minimal walking simulator game developed in Unreal Engine 4. The player begins on a collapsed bridge in the desert and must discover what is inside the bunker they find behind them. You can download it here for Windows.

Orderall

2018 — 2019

I founded Orderall to find a way for smaller restaurants to compete when so much food is purchased through platforms like Uber Eats and DoorDash. Orderall was a way for users to make orders by a simple web interface, and for businesses to take those orders via landline phone calls, bypassing the need for expensive equipment and high fees that cut into already thin margins.

Blog posts

Local LLM-powered NPCs with llama.cpp, Mistral7b and StyleTTS2

23/12/2023

There has been a lot of hype regarding models like GPT4, Claude and others. They're great, I use them a lot, but there are scenarios where using them might not be the best option. For instance, if you're making a game and want to have NPCs that can talk to the player dynamically, you probably don't want to have to send the player's message to a server, wait for a response, then send that response back to the player, since that would introduce a lot of latency, among other issues. I used llama.cpp with Mistral7b to generate the dialogue, and StyleTTS2 to generate voice lines, along with Unreal Engine 5 for the rendering.