Archana Ramalingam enhanced the nod-ai/SHARK-Platform by improving the stability and reliability of perplexity evaluation for language models. She addressed issues in the Perplexity class by refining how batch sequence lengths and cache states are handled, ensuring accurate metric computation. Archana also streamlined the continuous integration process by updating PR-triggered workflows, adjusting test parameters, and improving logging for better observability. Using Python, PyTorch, and YAML, she introduced dedicated test prompts and device adjustments to support robust CI testing. Her work delivered deeper test coverage and more dependable model evaluation, reflecting a thoughtful approach to both code quality and workflow efficiency.