Official repo for "The impact of internal variability on benchmarking deep learning climate emulators" in JAMES25 (public)

A Genetic Algorithm (GA) / Discrete Particle Swarm Optimization/ Hybrid (GA-PSO) for nuclear fuel optimization using ML surrogates (DNN, KNN, Random Forest, Ridge) and OpenMC. Optimizes fuel loading patterns for a target k-eff and minimal Power Peaking Factor (PPF).

Performance Modeling of Data Storage Systems Using Generative Models, IEEE Access, vol. 13, pp. 49643-49658, 2025, doi: 10.1109/ACCESS.2025.3552409

This repository provides an implementation of algorithmic support for dynamic pricing based on surrogate ticket demand modeling for a passenger rail company on open data.

A novel neural network for effective learning of highly impulsive/oscillatory dynamic systems by jointly utilizing low-order derivatives

Uncertainty quantification, Bayesian inference, and scientific ML for physical/biological models

Python code for running the numerical experiments in the paper "Neural Network Accelerated Implicit Filtering: Integrating Neural Network Surrogates With Provably Convergent Derivative Free Optimization Methods" by Brian Irwin, Eldad Haber, Raviv Gal, and Avi Ziv.

Unified GPU pipeline merging Taichi LBM (LES + FSI) and JAX to train real-time neural surrogates for unsteady bio-aerodynamics via continuous online learning.

Python Low-Code System Solutions

Master’s thesis project on AI-based crowd evacuation modeling using the SWIM algorithm, integrating simulation data and neural network surrogates.

Official code for "Data-Driven Prediction of Stress–Strain Fields Around Interacting Mining Excavations". Uses FEM data & Machine Learning (RF, LightGBM, CatBoost, MLP).

GNN surrogate for 2D shallow water simulations

EPDE - partial differential equations discovery framework

CARDIOKOOP - Control-aware Koopman deep learning framework for real-time hemodynamic forecasting and cardiovascular digital twin applications.

An attempt to bridge the gap between AI inference and wet-lab experiments. A practical loop to accelerate scientific discovery using small-sample data and expert knowledge.

Uncertainty Quantification pipeline for reinforced-concrete FEM (Abaqus) using surrogate models (PCA+GPR, AE+GPR, Shape-Scale) for fast Monte Carlo propagation, sensitivity analysis, and reliability metrics.

MeshGNN + PANN hybrid surrogate for hyperelastic structural analysis, deployed on HoloLens 2 — MSc thesis at RWTH Aachen