Solid mechanics simulations allow material engineers to craft materials at microstructure level. However, these simulations are based on non-linear physics which becomes enormously expensive to compute (using deterministic fixed-iteration approach) as the system size increases. This paves a way for data-driven learning based approach which offers the promise of speedup by orders of magnitude. We are employing deep learning models to test this hypothesis.
Tools used: DAMASK, TensorFlow, PyTorch
Poster presented at MLSS-2022.

As a part of week-long study excursion, I worked with an interdisciplinary team of master students on developing an algorithm to detect gravity wave events which are essential for reliable weather predictions. These waves were modelled as Morlet wavelets at given point, whose parameters were found using FFTs and non-convex optimization. The code was written in Python using NumPy, SciPy, JAX libraries.
Report

This semester-long lab project focused on building local-optima-tracking software for dynamic time-dependent functions. It was build on a global optima search routine developed by Dr. Jens Deussen and provides a switching criterion between global and local search. The software in written in C++ using dco/c++ library for automatic differentiation.
Report

Started as a bachelor's project at the makerspace of NIT Warangal, Project Aakaar has expanded into a global network of designers, thinkers, managers, and engineers, with the common goal of making technical education accessible in the special schools of developing countries.
Media Coverage   Report

As part of a master's course on iterative solvers, I programmed GMRES, CG and MultiGrid iterative solvers, along with Eigensolvers like Power Iteration and Lanczos method, all in Python. These were tested on big matrices taken from Matrix Market, and loaded in compressed row storage format.
Code

As a MITACS scholar, I worked on optimizing LPIM injection stage for metallic feedstock using FEM simulations and experimentation. I also worked on an initial layout of a new viscosity model that accurately captured viscosity behaviour for our application.
Publication

As part of a bachelor's elective on computational material science, I simulated precipitation in ternary systems to study growth and coarsening kinetics using Phase-Field-Modeling.
Code

As a summer research intern, I worked on synthesis and characterization of imine-linked covalent organic frameworks (COFs). These molecules were added into PDMS to form a composite, which exhibited self-healing property due to the reversible nature of the imine linkages.