Deep Neural Networks and Scientific Computing: Worked on improving training of ODE based formulation of recurrent neural networks (RNNs) by integrating attention mechanisms and attempting to provide theoretical guarantees for exploding/vanishing gradient problems with practical experimentation!

Adaptive Stochastic Algorithms for PCA: Developed a novel adaptive algorithm—a hybrid between vanilla Oja's method and Adam, viz ASAP, for the streaming PCA problem in the stochastic batch size setting.

Optimization and Quantum Computing: Worked on understanding iterative optimization methods that can achieve speedup on quantum Computers.

Low-rank Matrix completion: Worked on developing algorithms for Matrix completion and Matrix Rigidity problem over finite-fields GF(p) and real fields.