Iyed Mekki

Designed and built an agentic SAS-to-Python migration framework to translate legacy SAS ETL workflows into maintainable Python pipelines. Built a hybrid retrieval engine combining SAS syntax matching, semantic reranking, and GraphRAG-ready issue-to-fix knowledge links. Created a manually curated SAS-to-Python benchmark dataset with reference outputs, evaluating migrations through computational accuracy and CodeBERT similarity.

Developed a validation engine using ML models (Random Forest, XGBoost) to control the quality of structured and unstructured multi-sector data. Implemented anomaly detection and semantic checks via NLP (embeddings, tokenization, coherence rules). Designed a modular architecture with a REST API, integrated into ETL pipelines, and automated report generation.