I built an end-to-end LLM fine-tuning pipeline for domain-specific instruction following, covering document processing, synthetic instruction data generation, supervised fine-tuning, evaluation, and inference. I also implemented PEFT workflows with LoRA and QLoRA using PyTorch and Hugging Face Transformers.
For a battery research knowledge platform, I developed document-processing components that turn research papers into structured data for RAG and instruction-dataset generation. My research projects also include machine-learning models for early battery cycle-life prediction and physics-based and DFT modelling of electrochemical systems.
