Machine-learning

CycleGANs were used to synthesize new data from simulated data with lab test data distribution characteristics to address the sparse lab test data issue in my master thesis project

As part of my master thesis project, we developed our own novel architecture Hybrid Physics Informed Neural Network which works in tandem with a physics based flow solver to predict flowrates.

Bayesian Neural Network architectures MC Dropout and Bayesian Ensembling were used to quantify uncertainties in the restrictor predictions of the cabin air distribution system in my master thesis project.

Machine Learning works to optimize Air Traffic Management done for the 2-day ADS-B Hackathon at Toulouse

These projects were completed as part of the Udacity Deep Learning Foundations Nanodegree program using an assortment of neural network architectures like MLP, RNN, CNN and GANs.