llm-powered digital twins for urban mobility

Urban mobility simulation is useful, but it can be difficult to make interactive. A practical digital twin needs data ingestion, a simulation backend, and an interface that lets users ask useful questions without learning every implementation detail.

My work on Web-SUMO and related digital-twin projects focuses on that connection point. The goal is to make traffic simulation easier to query, modify, and evaluate with AI agents while keeping the simulation grounded in structured data.

The long-term challenge is reliability. LLMs can help users interact with a simulation, but the final decisions still need to come from constrained tools, explicit data, and measurable outputs.