Cities increasingly delegate consequential decisions to computational systems: how traffic is routed, which bridges and pipes are repaired before they fail, which neighbourhoods receive early warning of a flood, how scarce public services are allocated. Dr Sargiotis's research starts from the conviction that the resilience of urban infrastructure depends on treating the engineering of these systems and the governance of the data they consume as a single problem.
His work proceeds along three connected lines, which correspond to three book series with Springer Nature. AI and the Intelligent City treats urban planning and governance as continuous, data-driven decision processes, from land use and digital twins to the ethics of urban AI. AI and Green Infrastructure addresses mobility, energy, water, waste and construction as systems that can learn and adapt. AI for Human-Centric Cities asks whom intelligent infrastructure serves, examining public policy, health, well-being, inclusion and sustainable civil engineering. Further volumes address data governance, MATLAB for civil engineers, disaster resilience, construction automation, civil engineering and planetary sustainability.
Methodologically, fifteen of the seventeen books share a common form. Most chapters review the published literature, often including documented city initiatives, and then formulate AI frameworks as explicit mathematical objectives and layered architectures, most often combining forecasting models such as LSTM networks and graph neural networks with reinforcement-learning controllers, digital twins and federated, privacy-preserving learning, and ending in a human-in-the-loop layer. In most volumes fairness, privacy and explainability, and in several the energy cost of AI itself, are written into these formulations as constraints rather than added afterwards. Across these fifteen volumes, the result is a sustained programme of conceptual work on how AI could be designed for, and governed by, public institutions.
Since 2026 this programme has been complemented by empirical work released as open-source research software: a federated multi-agent controller for energy-aware public transport evaluated on the Boston MBTA network, and a leakage-resistant protocol for evaluating machine-learning classifiers. Both are archived with their data, so that every reported figure can be regenerated and checked.