Worked on the pytorch/executorch repository to implement a configurable learning rate feature for fine-tuning large language models. This enhancement allowed the learning rate to be dynamically sourced from YAML or JSON configuration files, enabling flexible experimentation without modifying code. Leveraging Python and PyTorch, the solution integrated seamlessly with existing llm_pte_finetuning workflows, supporting config-driven hyperparameter management. The approach improved training reproducibility and accelerated experimentation cycles, aligning with modern MLOps practices. No major bugs were addressed during this period, with the primary focus on delivering robust configuration management and enhancing the flexibility of machine learning model training through structured configuration parsing.