Machine Learning Researcher, Postdoctoral fellowship, University of Michigan
I designed and developed an open-source Python package for building predictive models adapted to engineering problems SMT which integrates continuous integration technology.
I led the development of the web-based airfoil design tool Webfoil, getting and cleaning the data, dimensional reduction, unsupervised and supervised learning. Interactive analysis of 2D airfoils compared to 1–2 hours using classical approaches.
I improved by 2.5% the prediction accuracy (relative error) of the aerodynamic performance of airfoils using deep learning (TensorFlow) techniques in multi-information source modeling.