Faustino Tchouwa

Managing the Data Quality System

Implemented an accurate component-tree algorithm in Python. Capable of implementing more than 100 filters. 60% reduction in signal-to-noise ratio while preserving 85% of an image’s edges. Implementation of a salt noise filter. Implementation of a filter from a deep learning model to reduce inference time by 70% while maintining the same accuracy

Made a rigorous extension of a PyTorch API for pruning, a technique to reduce the number of parameters for Deep Learning models based on different metrics. Reduction of 70% of our Deep Learning architecture such as InceptionTime or Encoder with still a strong accuracy of 92% on few datasets for Time Series Classification task, to deploy those models on smaller devices. Collaborated to write a research paper.