Shayan Dodge
About Me
I am a PhD researcher in Electrical Engineering at the University of Pisa, Italy, working at the intersection of Scientific Machine Learning, Computational Electromagnetics, and AI for Physics.
My research focuses on applying AI to electromagnetic simulation, power and energy systems, bioelectromagnetics, inverse problems, and scientific computing.
I develop Physics-Informed Neural Networks (PINNs), variational PINNs, neural operators, generative AI models, and deep learning frameworks for solving PDEs, surrogate modeling, signal analysis, optimization, and simulation acceleration.
Research at a Glance
| Category | Focus |
|---|---|
| Scientific Machine Learning | PINNs, Variational PINNs, Neural Operators |
| Deep Learning | RNNs, CNNs, Autoencoders, Generative AI, Residual Networks |
| Computational Methods | FEM, FDTD, BEM |
| Applications | Computational Electromagnetics, Power Systems, Bioelectromagnetics |
| Research Topics | Inverse Problems, Optimization, Surrogate Modeling, Scientific Computing |
| Software | PyTorch, TensorFlow, JAX, MATLAB, CUDA |
Key Contributions
Physics-Informed Neural Networks (PINNs)
PINNs combine machine learning with governing physical laws, providing a powerful framework for solving scientific and engineering problems. My research contributions in this area include:
INI-VPINN: A variational PINN framework with implicit treatment of Neumann and interface conditions for multi-material domains with complex geometries [Paper][GitHub]
A STacked Adaptive Residual PINN (STAR-PINN) for time-domain magnetic diffusion [Paper][GitHub][LinkedIn]
Hybrid Boundary Element–PINN Method for Electromagnetic Analysis [Paper][GitHub][LinkedIn]
Forecasting Lightning Effects in Electrical Systems (FELINES)
FELINES develops signal processing and deep learning methods for analyzing FDTD-simulated lightning electromagnetic signals and assessing lightning-related risks in power systems and transmission lines. Our contributions include:
More details are available on the project website and GitHub.
AI-Based Optimization of Transcranial Magnetic Stimulation (TMS)
This work develops a generative AI framework for transcranial magnetic stimulation (TMS) planning. A variational autoencoder (VAE) generates desired electric field distributions from clinician-defined targets, while a convolutional neural network (CNN) predicts the corresponding coil position and stimulation parameters using MRI-based head models and FEM simulations accelerated through model order reduction (MOR).
More details are available on the project website.
Recent News
[06/2026] Participated in ACES 2026 and CEFC 2026 in Thessaloniki, Greece, contributing to five works spanning Scientific Machine Learning, Physics-Informed Neural Networks, Computational Electromagnetics and Biomedical Engineering.
[05/05/2026] Joined the IEEE Journal on Multiscale and Multiphysics Computational Techniques (JMMCT) as a Student Editorial Assistant. Excited to contribute to the journal and further develop my research and editorial experience. [IEEE JMMCE]
[14/04/2026] Our paper, “PINN-based resolution of inverse non-linear magnetostatic problems,” has been published in COMPEL.