Weβre excited to highlight the work of Amna Hassan, a student in our research mentorship program, who recently completed a research project titled:
“Automated Unity Game Template Generation from GDDs via NLP and Multi-Modal LLMs”
In this project, Amna designed a system that leverages Large Language Models (LLMs) and Natural Language Processing (NLP) to automatically generate Unity game templates directly from Game Design Documents (GDDs).
She fine-tuned a LLaMA-3 model specifically for Unity game development using real-world GDD-code examples and developed a custom Unity integration tool, enabling game developers to go from design to implementation with just a few clicks.
π Highlights of the Project:
- A structured pipeline to parse GDDs into key game components
- Fine-tuned LLM that generates Unity C# scripts aligned with design specs
- A Unity plugin that handles script generation, dependency mapping, and documentation
- Comprehensive evaluation across multiple game genres and LLM baselines
Her fine-tuned model outperformed several leading LLMs in compilation success, modularity, and Unity best practices.
Amnaβs work is a great example of how students in our program take on meaningful technical challenges, learn advanced tools, and contribute to real-world innovation.
Leave a Reply