At Keywords Studios, I support LLM training and optimization using prompt engineering, supervised fine-tuning, and iterative refinement. I also design prompt-based tests and review model outputs to identify failure patterns and strengthen evaluation workflows.
I built an end-to-end RAG PDF chatbot with Python, LangChain, Ollama, ChromaDB, and Streamlit, including source snippets and page-level references for traceable answers. I’ve also developed a YOLOv8 CCTV weapon-detection prototype and a Power BI sales analysis dashboard.
