I build AI pipelines that turn sparse environmental observations into continuous, analysis-ready data. Working across radar products, satellites, and in situ networks, I use generative adversarial networks, sequence models, and physics-informed learning to sharpen spatial detail, recover missing values, and flag extremes. These approaches make high-resolution digital twins of urban climates possible, preserving physical structure while amplifying signal fidelity for downstream decision tools.
Current priorities include tailoring AI architectures to multi-sensor fusion, enforcing physical constraints during training, and delivering interpretable outputs for planners and scientists. By combining deep learning with process knowledge, I aim to deliver fast, trustworthy datasets that support early-warning systems and resilient infrastructure design.