Case study

Kick prototype: ML agents + multiplayer + zero-install

Built a Unity prototype combining ML-Agents gameplay decisions, deterministic multiplayer, and WebGL-ready instant play.

2024-12

Solutions Architect

ML Agents Multiplayer Unity WebGL

Working ML-Agents training loop for one-shot gameplay decisions

Deterministic procedural generation synced across multiplayer sessions

WebGL-ready workflow for instant playtesting

Unity ML-Agents C# WebGL Cloud Code

Kick prototype architecture flow diagram

Challenge

Validate whether ML-driven decision-making can feel skillful in a one-shot gameplay loop while keeping multiplayer sessions fast, deterministic, and zero-install.

Approach

Implemented a high-level ML-Agents decision layer (kick offset, power, yaw).
Built deterministic procedural generation with host-selected seeds.
Created a lightweight multiplayer session controller to gate readiness and timing.
Added WebGL-friendly tooling to enable instant access for playtests.

Results

Working ML training loop with stable reward shaping.
Deterministic multiplayer sync across sessions.
Browser-first workflow for rapid iteration.

Key metrics

Measured impact

Metric	Before	After
Decision layer	Manual tuning	ML-Agents policy
Session sync	Manual coordination	Deterministic seed sync
Playtest speed	Install required	WebGL instant access

Implementation notes

High-level decision layers trained faster than raw input imitation.
Deterministic generation simplified multiplayer sync.
Web-first testing reduced iteration friction dramatically.