• Researched and developed a chatbot based on the GPT-4 model to incorporate Motivational Interviewing (MI) strategies within generated dialogues to assist ambivalent smokers in making the decision to quit smoking • Prompt engineered GPT-4 to be 100% adherent to MI standards and to produce on topic outputs• Built a web page using JavaScript /HTML/CSS in frontend and JavaScript / Python in backend• Deployed web page on AWS Amplify to allow users on Prolific to interact with the chatbot

• Processed and utilized over 10 000 question-answer pairs as data points for experiment using Python• Implemented the PyTorch library to use various LLM such as GPT-2, GPT-3, and LLaMA 2 to produce the probability of each answer for each question • Implemented the BERT model to calculate the surprisal value for each question in the dataset

• Devised a google chrome extension, fortified with API calls to LLMs, capable of automating repetitive web data entry tasks• Researched various programming by demonstration systems, and developed a novel system that can capture an unlimited number of actions on a website, including 10+ distinct actions• Built and deployed the product using HTML / CSS on the frontend and JavaScript / Python on the backend to capture and repeat user actions on websites, achieving a 91.2% accuracy rate in automating the user’s intent• Managed a team of interns, overseeing project timelines, and delegating tasks (second author of published paper “DiLogics”)