Usman Tahir Qureshi

Master's

Computer Science

Grand Valley State University

Software Developer

-Collaborated with client teams to define specifications for a web-based microservice utilizing Large Language Models (LLMs) for summarizing peer review activities, resulting in a 30% reduction in review time. - Analyzed and curated peer review datasets to resolve inconsistencies, improving data quality by 25% using Pandas. - Optimized fine-tuning strategies for the mistral-7b language model with PyTorch, achieving a 15% improvement in summarization accuracy. - Conducted cost-efficiency analysis of Cloud GPU providers, that saved $5,000 annually on training costs. - Evaluated LLMs including Cohere, Gemini, and GPT-3.5, establishing a robust baseline summarizer. - Developed an Android application using Kotlin and Firebase, integrating Google Maps API for location-based services.