As part of my capstone project in the Master’s in Data Science program at the University of Virginia, this research evaluates various LLM evaluation frameworks, emphasizing bias detection, response quality assessment, and robustness testing. The study leverages multiple datasets and methodologies to benchmark state-of-the-art approaches for ethical and reliable AI assessment.
This blog post explores the three-stage bias detection process in DeepEval, an LLM-based evaluation system that quantifies bias in AI-generated text. The methodology leverages structured validation, templated prompts, and a scoring framework to assess bias across multiple categories.
As part of my GPU Architectures course, this project explores profiling and optimizing attention mechanisms in transformers using custom CUDA extensions. The focus is on reducing inference and training latency through kernel-level enhancements, improving GPU resource utilization for deep learning workloads.
This dataset contains all irregular verbs in known Romance languages.