Matthew H. Kutarna
Matthew H. Kutarna
Machine Learning Engineer in Vancouver, BC

About Me

I am an Engineer with experience in Data Science and Machine Learning and a background in text-to-speech (TTS) / voice cloning tools. I have experience in deep learning and am currently developing a tool to synthesize audiobooks from ebooks with voice cloning. I have interests in audio, robotics, and engineering applications.

My background is in Engineering Physics, with stops in robotics, electronics, and aerospace manufacturing.

I am interested in Machine Learning, Data Science, and Software Engineering roles - I'm looking forward to leveraging my unique skill set in a highly technical environment.

Highlights

  • Proficient in Python, MySQL
  • Frameworks: Pandas, Numpy, Scikit-learn, Pytorch, Tensorboard, Altair, Matplotlib, Tableau, Github Actions, Tensorflow/Keras
  • ML Techniques: Deep Learning, Text-to-speech , Voice cloning, Transfer learning, Dashboards, APIs
  • Software Engineering: Version control, Continuous integration, Scrum, Agile
  • Comfortable in CLI, Linux user, Docker
  • Past experience with C++, Java, R, LabView, ROS
  • Bilingual: English / French

Experience

Education

  • Bachelor's of Applied Science in Engineering Physics, 2013

    University of British Columbia, Vancouver

  • Certificate in Data Science, 2022

    University of British Columbia, Vancouver

Project: Audiobook Generator

  • Python / Streamlit web app: generates custom-voiced audiobooks from ebooks.
  • Pytorch / TTS: user provides ebook - or other document - and selects audio sample of desired narrator. The tool synthesizes audio (text-to-speech), then applies voice cloning techniques.
  • Development in progress: repo can be found on github, here: https://github.com/mkutarna/audiobook_gen

Project: Automatic Ticket Routing

  • WebRT / Best Practical: developing a tool which sorts 100+ incoming tickets per day, saving hours of additional guess work by staff.
  • ML-based sorting: applies NLP and custom trained model to classify tickets into appropriate destination queues.
  • Data scraping: pulled data from ~500,000 tickets via the WebRT REST API for training and testing.

Contact Me