Wael Lejmi

• Built a Text-to-SQL agent enabling natural language access to enterprise relational databases while preserving data sovereignty through private deployment. • Architected a deterministic LangGraph workflow to ensure auditable query generation optimized for self-hosted models . • Developed a RAG-based semantic layer combining schema context, business rules, and validated query history to improve SQL generation accuracy and business alignment. • Implemented human-in-the-loop SQL approval with editable query previews, alongside conversation state persistence. • Integrated support for 6 SQL dialects with automatic discovery of Docker containers, exposing the agent via Web UI, CLI, and MCP.

• Developed an AI validation agent to assess user-submitted project reports against ISO 14040 requirements using retrieval- augmented document analysis. • Fine-tuned a domain-specific NER model using LoRA to extract key lifecycle metrics from unstructured text.

• Built a Point of Sale application for retail using Flutter, AdonisJS and PostgreSQL, designed for order lifecycle management.