At Envisioned Strategy Consulting, I architected an enterprise Agentic AI chatbot that supports natural-language interaction with HRMS, Payroll, Attendance, CRM, Timesheet, and Staffing data across 124+ database tables.
I designed a 10-node LangGraph pipeline for query routing, schema retrieval, Text-to-SQL generation, validation, execution, and self-correction. Its schema intelligence layer retrieves only the relevant tables, columns, and relationships for each query.
I also built SQL security and governance controls with read-only execution, company-level isolation, RBAC-based access, and protection for sensitive employee and compensation data. A zero-token FastRouter and reflection loop helped reduce unnecessary LLM calls and recover from failed SQL.
On TheWorkNova, I optimized context construction and routing, reducing LLM API consumption by about 65% and achieving 94% SQL self-correction recovery in controlled scenarios. My other projects include a traffic signal optimization model and a Streamlit application for chronic kidney disease prediction.
