Machine Learning Developer in Bedford, United Kingdom

This methodology predicts the outcome of computational fluid dynamics (CFD) simulations through artificial neural networks. The input data is high-dimensional (tens to hundreds of parameters), and each data point is computationally expensive, taking hours to days to generate. The methodology employs advanced dimensionality reduction algorithms and function reformulation to efficiently train a neural network on a low-dimensional space that is then exploited for prediction purposes on unseen data.

This methodology helps exploit gradient information when constructing polynomial chaos expansions for uncertainty quantification. The sampling requirement is significantly reduced, enabling higher accuracy models. Exploited the method to perform robust design optimization of an axial fan for an aircraft engine.

BabyClub is an Android application for stage-by-stage pregnancy tracking, including useful utilities like weight management, contraction timer, and to-do lists. User management is handled through Google Firebase and the app monetization is implemented through AdMob.

My academic work focused on using data-centric reduced-order machine learning modeling to characterize the behavior of complex multidimensional systems. The work employed AI coupled with statistical learning to optimize the performance of an aerospace engine component.