Hey I'm Vedant. I am keenly interested in Coding and want to expand my skillset
• Researched diverse techniques to optimize downstream NLP tasks, including reducing hallucination and fine-tuning
an LLM under hardware constraints (PEFT).
• Automated data loading for RAG to support multiple documents of different types as input to an embedding model
to store in Vector Similarity Search Store (FAISS).
• Designed and executed proofs of concept (POC) using AWS EC2 on Quantized LLMs (Zephyr-7B-Beta &
Mistral-7B-Instruct) integrated with the RAG pipeline, achieving a 90% increase in performance on Generic and
Domain specific Datasets