Computational Building Scientist · Climate Resilience Researcher IntCDC Cluster of Excellence, University of Stuttgart
I am a Computational Building Scientist with a PhD in Architecture from the University of Seville (2026), an MSc in Landscape Architecture from the University of Tehran, and a strong background in computational modelling and urban microclimatology. I am currently an Associate Researcher at the IntCDC Cluster of Excellence (DFG Excellence Strategy), University of Stuttgart. I specialise in coupling high-fidelity physics-based simulation with machine learning to generate clear, actionable insights that support climate-resilient design decisions for buildings and urban systems.
My technical toolkit includes Python (PyTorch, scikit-learn, CatBoost, NumPy/pandas), multi-objective evolutionary optimization (NSGA-II/III, Optuna), and physics-based engines (EnergyPlus, ENVI-met, OpenFOAM, TRNSYS, Ladybug Tools). I build complete research pipelines — from climate data extraction and downscaling (CMIP6 / EURO-CORDEX, MODIS/Landsat remote sensing), through surrogate modelling, global sensitivity and causal analysis, to multi-objective optimization under future climate scenarios — delivered as reproducible open-source frameworks, validated simulation campaigns, and decision-support evidence for architects, engineers, and policymakers.
- Climate non-stationarity is the design problem , I quantify where historical design baselines fail (compounded UHI–heatwave forcing, retrofit compliance risk) as computable thresholds, not narrative warnings.
- Physics is ground truth; machine learning is the accelerator — surrogates are disciplined by simulation: sealed content-addressed campaigns, spatial no-leakage splits, uncertainty quantified by construction, claims validated back against the solver.
- Reproducible or it doesn't exist , every pipeline ends in versioned, hash-sealed artifacts with deterministic multi-seed protocols; figures must regenerate from the repo.
- Building-resolution metamodels of urban microclimate , a dual-head 3D-CNN emulating ENVI-met air and facade fields from ray-cast exposure geometry, validated against the solver's own radiative diagnostics (Szeged V-DEI); manuscript in preparation.
- Certified design rules for PCM-integrated envelopes , sensitivity, causal discovery and symbolic regression converged into transfer-limited design rules across eight climate zones; under review.
- Street-resolution future weather synthesis — EURO-CORDEX trajectories fused with heatwave detection and UHI intensification into validated 50 × 50 m EPW files.
- Longitudinal resilience of residential blocks (IntCDC, Stuttgart) — multi-objective optimization of energy, PV potential, and outdoor comfort across climate horizons for a representative Stuttgart district.
| Repository | What it provides |
|---|---|
| nepenthe-auto-sof | Auto-SOF — object-oriented framework automating surrogate modeling and multi-objective optimization (18 regressors, stacked ensembles, NSGA-II / DE, live Pareto exploration) |
| szeged-vdei-metamodel | Dual-head 3D-CNN metamodel emulating ENVI-met; ray-casting exposure features, sealed dataset manifests, full multi-seed protocol |
| pcm-causal-symbolic-framework | End-to-end pipeline: simulation campaign → surrogates → sensitivity, causal discovery, SHAP, symbolic regression → design rules |
| multiobjective-building-energy | Surrogate-accelerated NSGA-II across coupled energy / comfort / carbon objectives under historical and future climate scenarios |
- Peer review — 200+ reviews for leading venues in building physics and applied energy (Elsevier, Springer)
- Mentoring — supervision of 5+ master's theses in building physics and computational workflows; guest lectures on urban energy modelling and climate-projection-informed design
- Funded research — positions and projects under Germany's Excellence Strategy (DFG EXC 2120/1, IntCDC) and an EU Horizon doctoral consortium on urban heat mitigation
Open to research collaboration on climate-resilient buildings and urban systems, methodological exchange on physics-informed machine learning, and review of computational-building-science work. Best reached by email · ORCID · Google Scholar