Data Scientist and Developer in Ottawa, ON, Canada

Built the first ML system from scratch at Noibu. The system predicted the likelihood that web bugs could be resolved, performed a thorough data analysis on the data that showed eCommerce bugs, and came up with an ML model that learned what kind of bugs were most likely to be resolved. Later, I collaborated with our design team and built a feature into our product that highlighted resolvable bugs that could be worked on easily. I created the model, tied the algorithm metrics to business KPIs, built the back-end in Rust, helped design the front-end and built a feedback system for continuous training and deployment.

Created a neural network-based solution to automate a high-dimensional data set classification. Before my approach was developed, medical professionals labeled 15 column data set on blood cells in breast cancer patients. The data consisted of hundreds of thousands of samples with a lot of noise and required hours of manual labor to clean and label the data. In my approach, I trained an autoencoder that automatically classified the cells and removed them. The model was delivered as a software package that was easy to use and retrain. The process saved a lot of manual labor time and allowed the hospital to analyze data at a new scale that was not previously possible.

Worked with a networking company to predict device failure in a given period. The objective was to predict the likelihood of a device failure. The company used this prediction to estimate the number of devices and which ones that would fail in the upcoming month to help prepare accordingly. The model I built was a deep learning model based on N-beats. The architecture was made for this particular purpose and performance well in predicting failures a few months into the future. The model was deployed on AWS, and I built the entire MLOps pipeline with support from other engineers.