Summary:
Main engineer on A Deep World, owning all gameplay code. I built the enemy AI and the flocking system behind the underwater ecosystem, which runs 900+ agents at 60fps.
1. Game AI Engineering
Predators that hunt by sight and sound. They follow, investigate, or lunge to eat the player, and a Sonar mechanics lets them find you even when you’re hidden.
Under the hood: a behavior tree decides what the enemy is trying to do by priority, a state machine in our script handles transitions between modes (patrolling, attacking etc), and we use Utility Ai theory to score between patrolling and doing the Sonar mechanic.
Each predator is assembled from PredatorModules, so a new enemy is mostly a new combination of modules.

2. Multi-thread flocking system
Based on Craig Reynolds’ boids paper. The first version ran all the flocking math in Update on the main thread, which didn’t scale. I moved the calculations to Unity’s Job System with Burst, so it runs on multiple threads, and Update now only applies the results and moves the agents. That’s what gets it to 900+ agents at 60fps.
Beyond basic boids:
- Predator avoidance: flocks evade nearby threats aggressively, like fish fleeing a predator.
- Event reactions: flocks respond to game events, like scattering at a big sound.
- Designer-facing modules: behavior is built from drag-and-drop modules (cohesion, goals, containment etc) that can be swapped at runtime.

Summary:
I designed A Deep World around one goal: an atmospheric, scary, dialogue-free stealth game where you play a fast but fragile creature. I led a 4-person team through weekly playtests to a vertical slice in the first 4 months. And on October 2026 I put it in front of real players with a demo on Steam Next Fest, getting to 1000+ wishlists.
1. Prototyping and ideation
Visit to California Science Center
After experimenting with a few mechanic-based prototypes, my teammate and I visited the California Science Center for inspiration. Observing strange sea creatures turned out to be exactly what we needed. I became fascinated by organisms with reactive, tentacle-like filaments which led to the use of procedural animation.
Preproduction
First couple of weeks I built movement-focused prototypes and experiments to convey character emotions without dialogue.
Every single prototype tried to answer a design question like: Can I make a player feel empathy through animation and movement? Can we make a physics-based movement not frustrating?

2. One mechanic, many uses
I wanted the game to be accessible, so the player has one main mechanic: Call. Instead of giving the player new abilities, I made the same ability do different things depending on the situation.

Using the “Call” to communicate with your friend

Using “Call” to get the monster’s attention

Using “Call” to open gates
3. The Game feel
Audio, and movement carry the experience goals, so I tuned them through playtests.
Movement: the first controller was physics-based, and playtesters said it felt sluggish. The dash could fight the player’s momentum, and turning around was hard. I moved to a faster, more precise controller, which made the chase scenes work better.
Audio: we treated audio as another character. With FMOD parameters and effects, the music turns scary when the player is near danger, and everything muffles when they hide.