Lead and implement full machine learning pipeline from scratch - from data gathering/preprocessing to production on an embedded device- Developed the company's first production deep learning model used to automatically control the motor assistance of a smart E-bike which resulted in a smooth and automated bike ride- Designed the data transformation pipeline used to turn unstructured bike rides to clean and useable data- Analyzed users bike rides data which resulted in better decision making- Built a syntethic data generation tool that mimics real-life bike rides to help augment the dataset used to train the models- Implemented a library for fast and optimized inference with trained models on an embedded device using pure C

Team of Normand Mousseau, professorTopic: Prefactor calculation using the quasi-harmonic approximation- Implement a function for the prefactor calculation using Fortran- Test the simulations on a silicon system to make sure all results are valid- Optimize the computation speed

Team of Florence Marchi, professorTopic: Development of a micro and nano manipulation station- Develop the software allowing the control of a micro-tweezer the station using Labview- Test the software by conducting experiments on different bio-molecules- Analyse and present data collected during the experiments- Actively participate in the search of new functionality to add to the software

Team of Monique Bernier, professorTopic: Development of an interactive Atlas of ice probability- Implement the statistical models to predict the future ice probability in the Hudson Bay using Python- Create a server to host all the spatial data using geoserver- Develop an interactive web interface allowing to query the available data and generate future predictions using Javascript and JQuery