The Loong Game
Our UNSW Battlecode entry demonstrates symbolic (classical) AI through world modelling, planning, and tactical search. The case study follows the bot’s design and evaluation, with source code and tooling.
Read the case study ↗Visualisation, graphics, and AI agents
Our AI experience spans symbolic (classical) AI and machine learning, including training large transformer models and developing optimisers. We also build interactive visualisations, simulations, and rendering engines.
AI case studies
Pong AI
An interactive demonstration of neural networks and evolutionary optimisation. Train an agent over 100 generations, then play against it. The training and game run in your browser using Rust and WebAssembly.
Play Pong AI ↗
over|yonder Espalier-2
Microsoft’s TRELLIS.2 turns an image into a textured 3D model, and its 1536 setting needs more memory than a 16 GB card has. Espalier-2 tiles the model’s decoders and chunks its export so the same pipeline runs on an RTX 5070 Ti, with every change measured against Microsoft’s output.
Read the case study ↗Rendering and engines
OpenZT2
We’re rebuilding Zoo Tycoon 2 in Rust and Bevy using a clean-room approach. You supply the original game archives, and OpenZT2 reads them.
Visit OpenZT2 ↗
Advent
An upcoming Unreal Engine 5.8 game using Nanite, Lumen, MetaHuman characters and the Unreal Animation Framework.
UpcomingHave an AI problem, a model to train, or a simulation to build? Get in touch.
Let’s put AI to work. ↗ Sydney, Australia · Working worldwide