Urban biodiversity decline has become a critical challenge in rapidly urbanizing coastal regions, where dense residential developments often replace natural habitats and ornamental rooftop gardens provide limited ecological function (Oberndorfer et al. 2007; Wooster et al. 2022). This paper addresses the gap between mainstream rooftop design and biodiversity support by presenting a computational framework that integrates participatory artificial intelligence (AI), biodegradable fabrication, and material agency to transform ornamental rooftop landscapes into adaptive ecological micro‑infrastructures (Sanders and Stappers 2008; Tokac, Bruyninckx, and Vande Moere 2023). Set in Sai Sha, Hong Kong, a peri‑urban coastal area undergoing rapid residential development with fragile hillside and coastal ecologies, the study centers on rooftop environments characterized by intense solar radiation, wind exposure, drainage limitations, and fragmented planting systems. This research questions: how can we redistribute design authorship across non‑expert stakeholders, algorithms, biodegradable materials, and rooftop ecosystems through participatory AI workflows?

The study employs a mixed-methods action research design to test how non-expert participants can meaningfully influence complex ecological outcomes, addressing four core objectives: to study how non‑experts can contribute to computational ecological design via AI‑assisted processes; to test how sketch‑driven AI outputs translate to functional, fabricable habitat geometries; to analyse how biodegradable materials shape iterative design decisions over time; and to refine how Sai Sha rooftops can be upgraded to biodiversity‑supporting micro‑infrastructures. The developed workflow links human creativity, generative AI, digital fabrication, and site‑specific environmental constraints to produce adaptive prototypes.