I currently work as part of a team of researchers and engineers focused on building virtual telepresence. I have worked on many different parts of this research pipeline, from writing data capture software, to large-scale data processing, to implementing computer vision and machine learning research.

• Increased keypoint detection quality by 10% by making changes to the training pipeline for keypoint detection ML model (PyTorch)
• Made architectural updates to keypoint detection ML model (PyTorch)
• Designed and implemented a failure detection and retraining pipeline for our state-of-the-art keypoint and segmentation detectors, leading to a 20% improvement in predicted keypoint quality (PyTorch)
• Developed a speech to facial animation prediction model using Bi-LSTM (PyTorch)
• Developed Computer Vision-based metrics to evaluate quality of keypoint annotations (Python)
• Designed and implemented pipeline for early evaluation of data capture quality (Python)
• Designed and implemented post-processing steps which made capture data consumable for research. Formalized and automated data preprocessing. (Python, C++, Bash)
• Created monitoring and alerting solutions for fast detection of issues in data capture pipeline (Python, SQL, PHP)
• Created a data management system to support PII data captured during user studies (ReactJS, PHP)
• Created a data management system to support large scale data annotation (Python, PHP)
• Developed and upgraded 2D and 3D annotation tool for annotation of over one million data points (Javascript - ReactJS, PHP, SQL, Python, C++).
• Collaborated in the design of a user study capture process that has captured over a thousand hours of data.
• Led the development of multiple custom software solutions to help Research Assistants interact with research tools and software (C++, Python)

Worked with researchers, clinics, and corporations to develop software and hardware solutions to problems at the intersection of physical movement and machine learning.

• PROMPT Institute - Functional data analysis (FDA) software tool - Developed FDA tool to evaluate consistency of articulation data over multiple repetitions (Matlab).
• The Speech And Stuttering Institute - Gentle Onset Software tool - Gentle Onset therapy is used to treat people who stutter. This Neural Network-based detector for gentle onset in speech enabled patients to practice Gentle Onset therapy at home.
• PROMPT Institute - Eguana Python - Developed software for analysis of Electromagnetic Articulograph data.
• Music Mind Inc - Clarinet Monitor - Developed hardware and software for a real time visualization of tongue position while playing a clarinet.

• Co-founded and led a healthcare software startup from infancy to producing 300k yearly revenue
• Led technical team and grew the team from two part-time to three full-time developers
• Planned project roadmap and supervised the development of iOS, WatchOS, Web and Server Side projects
• Collaborated with hospitals and researchers in Toronto to produce technical innovations in health care by providing recording, hosting and analysis solutions for human kinematic and physiological data

• Consulted leading businesses and banks in Canada & USA on iOS strategy and mobile app design
• Developed iOS applications and supervised a software development team of 4 people to produce software using Agile methodologies
• Planned project roadmap and supervised the development of iOS, WatchOS, Web and Server Side projects
• Responsible for emergency bug fixes for multiple projects on Web, iOS, Android and BlackBerry platforms