• Researching and building benchmark for the novel screen off effect on uncertainty in temporal causality• Generation and evaluation of fine-grained defeasibility in causality for LLMs• Supervised by Prof. Boi Faltings

• Voluntary Commitment for Social Dilemmas in Multi-Agent Reinforcement Learning• Supervised by Prof. Pascal Poupart

• Achieved SOTA 98.75% audio classification acc. on ESC-50 by implementing a BEATs fine-tuning pipeline• Investigated semantic-rich tokenizers’ and video transformers’ efficacy for gait analysis• Researching diffusion and GAN based Image-to-Image Translations solution for speech enhancement

• Built an End-to-End AWS deployment parallel processing pipeline to consume 1,000+ 2hr videos daily from 13 models with <10 EC2 Instances• Implemented an automated annotation pipeline by researching Community Detection, Active Learning, and Positive-Unlabelled Learning• Implemented an automated annotation pipeline with Community Detection and Active Learning

• Developed a Brain-Computer Interface API involving action-classification via EEG signal analysis• Deployed CCA with KNN and IT-CCA to classify neural activities that reached 99.7% accuracy• Built a Socket API that hosts models, retrieves the data, and returns the predicted outcome

• Ranked 3rd domestically in Skills Challenge, Canada; Qualified for VEX Robotics World Championship ’20• Programmed driver and autonomous events in C++ and simulated robot designs in SOLIDWORKS