Michele Pratusevich

Title: Michele Pratusevich: Innovator in Machine Learning Model Approximation

Introduction

Michele Pratusevich is a notable inventor based in Somerville, MA (US). She has made significant contributions to the field of machine learning, particularly in the area of model approximation. Her innovative work has the potential to enhance the efficiency of machine learning applications.

Latest Patents

Michele holds a patent for a system titled "Trained Model Approximation." This invention allows for the creation of an approximated model from a trained machine learning model, such as a neural network. The approximated model operates using fewer computing resources than the original trained model. Notably, this system can generate the approximated model without the extensive training data required for the original model. It relies on data that describes the trained model and an indicator for how closely the approximated model should correspond to the original. Various lossless and lossy approximations can be performed to produce multiple approximated models, which can be utilized during runtime operations to achieve significant speed and cost savings.

Career Highlights

Michele is currently employed at Amazon Technologies, Inc., where she continues to develop her innovative ideas. Her work focuses on improving machine learning technologies and making them more accessible and efficient for various applications.

Collaborations

Michele collaborates with Jeremy Samuel De Bonet, contributing to the advancement of their projects and sharing insights in their field of expertise.

Conclusion

Michele Pratusevich is a pioneering inventor whose work in machine learning model approximation is paving the way for more efficient computing solutions. Her contributions are significant in the realm of technology and innovation.