physics-ml

ReservoirNeuralBench – A controlled benchmark of neural surrogates for 3D reservoir simulation on the Norne field.

This repository contains Jupyter notebooks from my learning journey with Physics-Informed Neural Networks (PINNs). These are not complete projects but serve as educational resources to explore core and intermediate concepts in applying neural networks to solve partial differential equations (PDEs).

Fine-tuned DeepCFD UNet surrogate model for CFD flow field prediction

Physics-aware machine learning study for event classification in high-energy physics.

Predict steady-state 2D laminar flow fields in milliseconds using a deep learning surrogate model for Computational Fluid Dynamics.

The missing middleware between LLMs and CAE.

FNO-RC: Fourier Neural Operator with Conformal Residual Coupling for PDEs

Does an energy-trained transformer encode entanglement structure it was never trained on? Pre-registered study with exact-solvable ground truth. Apparatus and instrument validated; measurement deliberately not yet made.

Active Learning–Guided Adaptive ESR Acquisition | Poster presented at ICESR 2026, IISc Bengaluru

Mechanistic interpretability of classical neural networks trained on quantum data — using sparse autoencoders to detect quantum structure in classical representations.

Physics-guided ML for tool wear classification from machining vibration signals using Random Forest and FFT features

σFlow-PDE: A drop-in H-Bar training engine that escapes the σ-trap in neural PDE solvers via live σ/δ/α ODE integration, autonomous phase curriculum, and auto-falsification.

This repository contains a lightweight CNN-based approach for top-quark jet classification, using data from CERN’s public Zenodo dataset.

Physics-grounded RL: a Gymnasium environment where agent safety is an attractor in the dynamics, not a reward penalty. 12D manifold with Lagrangian + SU(2) gauge mechanics. CPU-friendly, CUDA-free. AGPL-3.0 + commercial.