Lawan Kurbah

Full Stack Engineer building .NET and React systems with AI/LLM automation for insurance and enterprise workflows. Delivers backend services, premium calculation engines, and conversational tools that improve performance, cut deployment complexity, and reduce LLM operating cost. Bridges product delivery and applied AI across simulation, analytics, and workflow automation.

Full Stack Engineer (.NET & React) / AI Engineer at CGI (2023-07 – Present)

• Built a high-performance premium calculation engine using Go and.NET Core for large-scale actuarial computations.
• Developed dynamic simulation workflows, improving system performance by 35%.
• Migrated services to.NET Core, reducing deployment complexity by 25%.
• Integrated RabbitMQ for scalable asynchronous processing across backend workflows.
• Implemented an end-to-end keyed-table comparison solution with API and UI integration to compare Excel-based keyed tables against database data.
• Developed an actuarial dashboard with ML-based forecasting using Python, Streamlit, and XGBoost.
• Reduced manual reporting effort by 40% and improved prediction accuracy by 30%.
• Built using Python, Streamlit, Plotly and XGBoost, enabling real-time simulation comparison and premium forecasting for large insurance dataset.
• Designed MCP-based orchestration with LangChain, implementing structured agent reasoning, dynamic tool selection, and graph-based execution.
• Implemented agent reasoning and dynamic tool selection using LangChain with structured prompts and graph-based execution.
• Developed a context-aware memory system using Chroma and integrated Azure OpenAI Service for scalable, multi-session conversational experiences.
• Contributed to a multi-agent backend using Python, FastAPI, and LangGraph for enterprise AI systems (EMMA, Ratabase360).
• Built a LangGraph-based disambiguation workflow that handles ambiguous inputs, pauses for clarification, and resumes execution seamlessly.
• Engineered a PostgreSQL (asyncpg) sliding-window context manager, cutting LLM token usage and API costs by ~60% while preserving conversation history.
• Exposed 10+ internal enterprise APIs as LLM tools using MCP over SSE, enabling secure AI-driven API access.

Bachelor of Technology in Computer Science & Engineering – Indian Institute of Information Technology, Kota (2019-08 – 2023-07)