Three pillars, one quality-engineering platform
A consolidated 6-paper portfolio that turns repository signals, LLMs, and runtime DOM analysis into measurable testing outcomes.
Predictive Analytics for Software Quality
Process-metric-based defect prediction at 296,457-instance scale across five mature OSS systems. Random Forest AUC 0.8998; cross-repository transfer F1 0.6306.
Intelligent Automation Architecture
Self-healing locators via tree-ensemble ranking on 2,400 mutation events; LLM test generation with deterministic rule verification across three domains.
Research-to-Production Translation
Risk-based test prioritisation with the gb-paper1-v4 model in a FastAPI microservice on GitHub Actions. Top 10% of files capture 43.82% of defects (4.37× lift).
Six papers, one program
Tap the title for details — or jump straight to the full list.
Process-Metric-Based Defect Prediction at Scale
Six classifiers · 296,457 instances · 5 OSS repos · RF AUC 0.8998
Cross-Repository Defect Prediction (LORO)
Leave-one-repository-out · Cross AUC 0.867 · F1 0.631
Risk-Based Test Prioritisation
Top-10% files capture 43.82% defects · 4.37× lift · CI/CD pattern
Self-Healing Test Automation
Tree-ensemble locator ranking · 2,400 mutation events · 7 refactor classes
LLM-Based Test Case Generation
312 requirements · 3 domains · verified with symbolic mutation indicators
Post-Execution Defect Attribution (Vision)
Fusing test-failure signals with SHAP-explained repository priors