During my internship at DataInsight, I worked on an innovative project to develop an AI Companion that could converse fluently in Latin/Roman Urdu with a focus on delivering sarcastic responses. The project entailed creating a custom dataset, fine-tuning various language models, and addressing numerous technical challenges to achieve a functional conversational AI.Key Accomplishments:Dataset Creation:Developed a custom dataset of over 700 Q/A pairs in Latin/Roman Urdu to facilitate model training.Model Fine-Tuning:Explored and fine-tuned multiple language models, including 32-bit, 8-bit, and 4-bit models, to improve their performance in Latin/Roman Urdu.Utilized quantization techniques to manage computational resources while maintaining language proficiency.Utilization of Computational Resources:Leveraged Kaggle’s free GPUs, addressing challenges in parallelization and resource allocation.Model Performance Analysis:Conducted thorough testing of various models, identifying the strengths and weaknesses of each in terms of language capabilities and knowledge retention.Deployment and MLOps:Highlighted the importance of MLOps in building and deploying AI applications, ensuring smooth integration and functionality.

Thoroughly researched and planned numerous events.

As the Department Head of Computing (Data Science) Events, I was responsible for conducting our flagship events namely, DataQuest and DataViz.I headed a team of 15 people that helped me successfully conduct the two events.