Projects (During my Masters and Undergraduate)
1. Using neural networks with domain decomposition to solve partial differential equations (Masters Thesis @TUM)
- Masters Thesis Report and Masters Thesis Presentation
- Used Extreme Learning Machine (ELM) method to solve partial differential equations.
- Focused on improving solution accuracy by employing better initialization techniques for neural networks, specifically utilizing the Sampling Where It Matters (SWIM) method and domain decomposition.
2. Domain decomposition to accelerate learning of Physics Informed Neural Network (PINN) (Seminar course @TUM)
- Seminar Paper
- Replicated a paper implementing Finite Basis Physics-Informed Neural Network (FBPINN). link
- Implemented overlapping subdomains to solve partial differential equations using Physics-Informed Neural Networks.
- Computed the final solution through additive schwarz domain decomposition.
3. Improve Building Efficiency for a Better Future (BGCE @TUM & Siemens)
- Github
- Collaborated with a team to develop an application to enhance building energy efficiency.
- Created a building’s thermal model and heating control based on user-provided floor plans.
- Contributed to this project as part of the honors program at BGCE, supervised by Siemens.
4. Simulating Free Surface Flows using Marker and Cell method (CFD Lab @TUM)
- Github
- Developed an object-oriented 2D parallel CFD solver in C++ for solving incompressible Navier-Stokes equations using Finite Difference Method (FDM).
- Extended the solver to support free surface flows utilizing the Marker and Cell Method, successfully simulating dam break and tank sloshing scenarios.
5. Entropically Damped Artificial Compressibility using Compact finite-difference scheme (Undergraduate Thesis @BITS)
- Undergraduate Thesis Report
- Utilized the EDAC technique to obtain solutions for the Navier-Stokes equation, bypassing the Pressure Poisson equation by implementing a pressure-evolving equation.
- Learnt about openMP and FORTRAN programming