KAJAL explores the use of Large Language Models to automatically infer the underlying grammar and structural rules of source code directly from software repositories. Instead of relying solely on manually written language specifications, the framework learns syntactic and semantic patterns that support program comprehension, software documentation, and automated software analysis.
The project investigates how AI can transform unstructured source code into structured representations that enable downstream software engineering tasks such as code analysis, documentation generation, repository understanding, and intelligent developer assistance. KAJAL contributes toward making software systems easier to analyze and maintain using generative AI.