Brandon Jaipersaud Email & Phone Number

Toronto, ON, CA

Toronto, Ontario, Canada

Using AI alignment and interpretability strategies from recent literature to mitigate implicit gender bias when using LLMs for story generation.Supervised by Frank Rudzicz, Tegan Maharaj, and Elliot Creager

Toronto, Ontario, Canada

Helped the Vector AI Engineering team with a 3-day industry workshop on Fairness in LMs. Made Jupyter notebooks enabling participants to perform various tasks such as fine-tuning Llama2 on the Vector cluster.

Toronto, Ontario, Canada

Led a research project from inception to publication where we develop a question-answering system for the Piazza discussion board using LLMs. We deployed our system across various Computer Science courses at UofT. Our work resulted in a publication at AIED 2023 where I was the first-author.Supervised by Jimmy Ba, Michael Zhang, Lisa Zhang, and Andrew.

Toronto, Ontario, Canada

Building image classification models to detect special vehicles and count vehicle occupancy which will be used in InvisionAI’s pipeline for applying penalties/exemption rules for vehicles on HOV/HOT lanes.

Designed survey questions and an interview script to gauge the effectiveness of a new teaching method that aims to use digital storytelling to deliver lecture content in a second-year CS data structures course (CSC263). Surveys and interviews were delivered to ~300 students in the Winter 2023 offering of CSC263.Supervised by: Andrew Petersen, and Tingting.

TA for 5 different CS courses:CSC384: Intro to AICSC311: Intro to MLCSC263: Data Structures and AnalysisCSC373: Algorithm Design and AnalysisCSC324: Principles of Programming LanguagesTasks include: holding weekly tutorials and office hours, designing and marking labs, assignments and exams, and discussion board monitoring.

Implemented a question-answering bot for the Piazza discussion board using the T5 transformer and the HuggingFace API. Our bot was deployed in the Fall 2022 instance of an Intro to Programming Class (CSC108) consisting of ~1000 students.Supervised by: Lisa Zhang and Andrew PetersenGithub Repo: https://github.com/brandonjaipersaud/qabot-t5

Toronto, Ontario, Canada

Used statistical tests such as mutual information, ANOVA and chi-squared to determine what factors influence whether an answer on an educational Q\&A discussion board will be marked as “helpful” by the question asker. Once these factors were discovered, we built a random forest model to predict whether an answer will be marked as “helpful” by the question.

Toronto, Ontario, Canada

Built an interface to popular ML frameworks such as Sklearn and PyTorch, enabling data analysts within the company to easily build and train ML models by simply configuring the parameters in a JSON file. I also created ML model and use-case documentation which will be used to determine an appropriate modeling strategy for company data.