Designed and evaluated a retrieval-augmented generation (RAG) pipeline on the FiQA financial QA dataset, benchmarking multiple retrieval strategies (sparse + dense) using MRR and Precision@K to optimize answer relevance and retrieval quality.

Fine-tuned a GPT-2 language model using LoRA on a customer support dataset, improving token-level and sequence-level generation quality through parameter-efficient adaptation and evaluation.