physics-informed-ai

Hohmann transfer orbit modeling with Hall-Effect thruster integration, implemented in Python, C++ and MATLAB with 3D visualization and validated delta-V calculations for LEO to GEO missions

A Physics-Informed Neural Network (PINN) pipeline that demonstrates the Dark Energy inverse problem and simulates the Quantum Vacuum Catastrophe.

Hybrid Physics-Informed Neural Network for Robot Joint State Prediction and Parameter Identification

Fourier-based Low-Frequency Spectral Dynamics Network for differentiable spectral learning, early-warning analysis, and physics-aware modeling of complex dynamical systems.

Beam Weaver is a learned-kernel Monte Carlo framework for photon transport in water.

neural-networks

SolarChain: Blockchain for Sustainable Energy Optimization

Predictive 3D Digital Twin and Physics-Informed AI system for real-time health monitoring and anomaly detection in MALE UAV aero-engines. Built for SIH 2026.

Physics-informed image denoiser built from scratch — Poisson-Gaussian noise modeling, residual CNN, no pretrained weights or APIs

Modeling the Tumor Microenvironment (TME) as a many-body gravitational system. A synthetic data engine that applies astrophysics principles (N-Body dynamics) to simulate immune-tumor topology and generate ground-truth spatial data.

A flexible PyTorch Lightning framework for climate forecasting, featuring a customizable, knowledge-aligned latent diffusion model for high-resolution predictions.

Physics-informed AI system for predicting aircraft bird strike risk using impact force calculations and machine learning techniques.

Universal foundational architecture specification for the Physical Matrix Neural Network (PMNN) superseding the Transformer paradigm.

Physics-Anchored Deepfake Detection Framework

Official PyTorch implementation of MFPIT (Multimodal Flood Physics-Regularized Transformer) for water occurrence mapping in the Indus Delta. Fuses multimodal Earth observation (Sentinel-1/2, MODIS) and climatological forcing variables under a soft hydrological mass-balance constraint.

Architecting stateful Agentic AI, aerospace simulations, SiN chip design workflows, and specialized deep-tech/medical LLM data pipelines

A five-lecture course building an SO(3)-equivariant neural network from first principles—rotation symmetry, irreducible representations, Clebsch–Gordan couplings. NumPy and SymPy only, no deep-learning framework. Notes, runnable notebooks and solutions, released weekly.

SymRegLab — Interactive Symbolic Regression & AI Equation Discovery Laboratory. 6 modules: GP evolution, physics rediscovery, expression trees, Pareto front, dimensional analysis, benchmark arena. Zero dependencies.

Predicting traffic accident risk using Physics-Informed Feature Engineering and Advanced Stacking Ensemble (XGBoost, LightGBM, CatBoost).