👋 About Me (CV)
I am an Italian-Brazilian PhD student at King Abdullah University of Science and Technology (KAUST) and a Doctoral Researcher with the industry-funded DeepWave Consortium, which builds machine- and deep-learning workflows for subsurface reservoir characterization. My advisor is Prof. Tariq Alkhalifah.
💻 I focus on deep generative AI — especially video diffusion models — to model multi-phase subsurface fluid-flow dynamics. I build latent conditional and progressive autoregressive video diffusion models tailored to reservoir simulation data, generating video sequences of key field variables such as CO₂ gas saturation and pressure build-up for geological carbon sequestration studies. My workflows rely on CUDA-aware, reproducible CPU/GPU parallel pipelines running at scale on the KAUST IBEX supercomputing cluster.
🔍 Research interests: Deep Learning · Deep Generative Modeling · Diffusion Models · Reservoir Simulation · Petrophysics & Well-Logging · Geological Carbon Storage
🔥 News
- 2026.06: 🎤 Presented Latent Autoregressive Video Diffusion Models for Fluid Flow Simulations at the 87th EAGE Annual Conference & Exhibition.
- 2026.01: 🎉 LAViG-FLOW: Latent Autoregressive Video Generation for Fluid Flow Simulations is out on arXiv, with code on GitHub.
- 2025.08: 🎤 Presented Towards Generative Modeling of CO₂ Geological Storage with Latent Conditional Diffusion Models at the Fifth International Meeting for Applied Geoscience & Energy (SEG/AAPG).
- 2025.06: 🎤 Presented A Conditional Diffusion Model for CO₂ Monitoring and Forecasting in Heterogeneous Geological Formations at the 86th EAGE Annual Conference & Exhibition.
- 2024.01: 🎓 Started my PhD in Earth Science and Engineering at KAUST, joining the DeepWave Consortium.
📝 Publications
LAViG-FLOW: Latent Autoregressive Video Generation for Fluid Flow Simulations
arXiv preprint, 2026
Vittoria De Pellegrini, Tariq Alkhalifah
[PDF] [Code]
- A latent autoregressive video generation diffusion framework that explicitly learns the coupled evolution of saturation and pressure fields.
- Dedicated autoencoders compress each state variable; a Video Diffusion Transformer models their temporal distribution.
- Autoregressive fine-tuning enables extrapolation beyond the observed time window, running two orders of magnitude faster than traditional numerical solvers.
A Conditional Diffusion Model for CO₂ Monitoring and Forecasting in Heterogeneous Geological Formations
86th EAGE Annual Conference & Exhibition, Vol. 2025, pp. 1–5, June 2025
Vittoria De Pellegrini, Damir Wamriew, Tariq Alkhalifah
[DOI] [Code]
- A conditional diffusion model for monitoring and forecasting CO₂ plume migration in heterogeneous geological formations.
- Latent Autoregressive Video Diffusion Models for Fluid Flow Simulations, Vittoria De Pellegrini, Tariq Alkhalifah, 87th EAGE Annual Conference & Exhibition, 2026
- Towards Generative Modeling of CO₂ Geological Storage with Latent Conditional Diffusion Models, Vittoria De Pellegrini, Tariq Alkhalifah, Fifth International Meeting for Applied Geoscience & Energy (SEG/AAPG), Expanded Abstracts 44(1), 1290–1294, 2025
- Development of Supervised Machine Learning Models for the Prediction of Well-Logs & Application on Wells at São Francisco and Santos Basins, Brazil, Vittoria De Pellegrini, M.Sc. Thesis, Politecnico di Torino, 2023
📖 Educations
- 2024.01 - 2027.12 (expected), Ph.D. in Earth Science and Engineering, King Abdullah University of Science and Technology (KAUST), Thuwal, Saudi Arabia. GPA 3.89
- 2020.10 - 2023.07, M.Sc. in Petroleum Engineering, Politecnico di Torino, Turin, Italy. Grade 108/110
- 2017.10 - 2020.07, B.Sc. in Civil Engineering, Università degli Studi di Padova, Padua, Italy. Grade 110/110
💻 Experience
- 2024.01 - present, Doctoral Researcher, DeepWave Research Consortium, KAUST, Thuwal, Saudi Arabia.
- Build deep generative models for multi-phase subsurface fluid-flow dynamics, including latent conditional and progressive autoregressive video diffusion models for reservoir simulation data.
- Run large-scale training and inference on the KAUST IBEX multi-GPU cluster with CUDA-aware job scripts and CPU/GPU parallel workflows.
- 2023.02 - 2023.07, Graduate Researcher, Politecnico di Torino, Turin, Italy.
- Developed supervised machine learning models for conventional and advanced well-log prediction, applied to wells from the São Francisco and Santos basins in Brazil using PETROBRAS datasets.
- 2022.09 - 2022.11, Geophysics Intern, Aramco Overseas Company B.V., Delft, Netherlands.
- Built a 3D multi-physics synthetic model for a geothermal site in Saudi Arabia, generated surface wave dispersion curves and produced synthetic seismograms.
🌍 Languages
Italian (bilingual) · Brazilian Portuguese (bilingual) · English (advanced) · Spanish (intermediate) · Khaliji Arabic (basic)