From Manual Labor to Infinite Scale

The Challenge: The Manual Bottleneck

The company runs a global PVP (Player vs. Player) coding challenge where students submit code to control robots in a 3D simulation. For two years, this was an operational nightmare.

• The Prep Time: To run a single match, the team had to manually convert and sync user code. It took 5 minutes of manual work per record.
• The Sync Trap: The biggest technical hurdle was latency. Running the user's code on the web while trying to sync it with a Unity simulation created massive delays. The "replay" never looked like what the user actually coded.
• The Risk: The process was unscalable. As participation grew, the manual overhead threatened to crush the competition.

The Solution: Full-Stack Automation

I took ownership of the project with a single goal: Zero Manual Touch.

1. Ingestion-Time Processing (5min → 1sec)

The bottleneck was processing code at runtime. I shifted the architecture left using Node.js.

• The Fix: I built an ingestion pipeline that converted and compiled the user's code the moment they hit "Submit."
• The Result: Instead of 5 minutes of prep time per match, the system simply fetched the pre-processed data. Retrieval time dropped to 1 second.

2. Browser-Based Parallel Execution

To solve the "Sync Trap" (where the code lagged behind the simulation), I moved the execution context.

• Innovation: Instead of trying to sync a server execution with a client view, I architected the solution to run both players' code instances in parallel directly in the browser environment.
• Impact: This eliminated network latency and server-side processing lag. The simulation became deterministic and perfectly synchronized.

3. The "Visual Rendering" Farm

Headless browsers could not faithfully render the complex 3D Unity environment for recording.

• The Infrastructure: I architected a dedicated farm of Windows Servers connected to active monitors (GPU-accelerated).
• The Workflow: A Node.js orchestrator dispatched matches to these servers. The servers ran the simulation 3 times, auto-captured the screen using standard recording tools, and generated the final MP4 highlight reel.
• The Future: This architecture laid the groundwork for a live broadcast system, capable of streaming match results in real-time to participants.

The Outcome

• 300x Efficiency Gain: Reduced match preparation time from 5 minutes to <1 second.
• Scalability: The system could handle thousands of submissions without requiring a single human operator.
• Visual Fidelity: By running on dedicated Windows hardware with GPUs, we ensured the recorded video was pixel-perfect compared to the user's code execution.