Project Sirius

 NOTE: Page under construction! More details & postmortem of my experience coming soon ☺

The project code name was announced here in a 2022 CDPR tweet.

Game development is a never ending learning journey; technology is always improving, and there is always a new technique to discover. I am so thankful for the incredible amount of knowledge that I gained at The Molasses Flood and for all the teammates who mentored me. 

I implemented custom behavior tree decorators in C++, such as semi-random subtree injection and pre-execution branch cooldowns. 

I planned & implemented multi-character managers for NPCs and enemies.

I collaborated with designers to create tools for tuning combat, adding injectable behaviors, and adjusting pathing patterns.

I extended and utilized many AI systems, both UE5 and proprietary, such as: perception, 2D and 3D movement, nav mesh generation, EQS, and spawners.

 One of the greatest challenges of AI, especially when using utility/score based solutions for pathing and other decisions, is debugging. Visualizations and data logging are vital for team members to understand why a character is making a decision, moving in a direction, etc. I created new Gameplay Debugger categories for our proprietary AI systems.

When I started working on the project, our player and AI used a proprietary action/effect system for all their behaviors. The decision was made a few months later to transition to Unreal's Gameplay Ability System. I learned a lot from comparing the workflows of both systems!

I generated faction-specific zones within larger maps using proprietary PCG systems and ensured that AI could query and respect the boundaries of these zones.

I extended nav mesh generation to support multiple meshes for the same agent type and learned about Sparse Voxel Octrees.

I created custom multi-AI spawners which selected spawn positions using the Environmental Query System.