Colin He

Fine-tuned and deployed a RoBERTa-flavoured NLP classifier using a mixture of hand-labelled and weak-labelled data, which scores thousands of text samples daily from Google Cloud's AI Platform via HTTPS. Researched and authored a front-facing blog post tutorial on weak labelling, a powerful semi-supervised dataset creation technique, based on experience with an internal project utilizing the Programmatic Python library. Explored SOTA methods for utterance analysis, including transcription, question detection, zero-shot classification and multi-language models (fasttext, Hugging Face, GPT-3). Developed an in-house Slack + Google Sheets solution for collecting high-quality human labels for large text datasets. Tech used: Python, Conda, Google Cloud, Pandas, Looker, Google Colab, Hugging Face, Snowflake

Responsible for automating various software development features, including testing and progress management. Worked directly with a cutting-edge web application used for cryptographic analysis and post-quantum cybersecurity use, alongside a team of highly motivated experts. Tech used: TypeScript, React, NodeJS, Docker, PostgreSQL, Selenium, Gitlab CI/CD

Constructed a real-time question-answering system for the course forum board using machine learning and MongoDB that provides instant feedback on academic integrity violations and saves tens of hours per week on student response duties, with modular design that allows for infinite configurability. Provided general course help for a 1500-student first-year CS course, including holding one-on-one help/debugging sessions with students and designing unit tests for assignment marking purposes. Performed various system administration duties in the course backend and server. Tech used: Python (TensorFlow, Scikit-Learn), Git, Bash, MongoDB, MySQL, Selenium

Involved designing and developing tools for automation of creation and importing of educational material using Python, as well as additional manual content creation, to be used in a cloud-based learning platform for STEM-focused high school and university students.

Basically Overcooked but in real life

Involves advertising the Port of Vancouver, communicating with the general public and performing miscellaneous behind-the-scenes duties for the Port of Vancouver.

Involves working with Microsoft Excel (2016) to display and organize data, communicating with higher-ups in a professional environment, interpreting instructions, and arranging presentations to be displayed in meetings.

Double Major in Computer Science (AI option) and Combinatorics & Optimization, cumulative average of 94%.