Urjasoft - Software, AI & SaaS Engineering
Founded 2016 • 150+ Projects • 14 Countries
Prototype exploring AST static analysis and LLM-assisted test generation.
An experimental architecture evaluating whether combining AST parsing with LLM reasoning can synthesize deterministic regression suites without manual authoring.
Legacy enterprise software architectures often suffer from insufficient unit test coverage, making refactoring, dependency upgrades, or framework modernization high-risk and labor-intensive.
Coupling deterministic AST-based boundary detection with LLM reasoning agents can synthesize executable, deterministic PHPUnit assertions without hallucinated APIs.
Multi-stage execution model separating probabilistic reasoning from deterministic state mutations.
nikic/PHP-Parser parses source files into Abstract Syntax Trees, mapping methods, parameters, and branch conditions.
Structured AST context formats method contracts and parameter boundaries for model reasoning.
Candidate PHPUnit test cases are generated following strict project typing and assertion rules.
Generated test classes are executed in an isolated runner to filter syntax and assertion failures.
“Observed Finding: In initial test runs, unconstrained LLMs generated plausible-looking assertions against non-existent method signatures. Supplying structured AST nodes directly eliminated method hallucination, though semantic assertions still require manual developer verification.”
Findings from this research area inform our engineering capabilities in enterprise software development.
An exploratory agent framework testing whether enforcing JSON Schema validation gates and cryptographic approval tokens between LLM decision steps prevents unintended API mutations.
An exploratory document processing pipeline evaluating whether preserving 2D coordinate bounding boxes during OCR improves key-value pair extraction from complex layouts.
An analytical research initiative evaluating whether lightweight statistical models can outperform simple moving averages on sparse, seasonal eCommerce catalog purchase data.
Our engineering team can evaluate your technical roadmap, suggest architecture patterns, and partner with you on high-assurance system execution.