AI Architect
& DATA ANALYST
Building custom Large Language Models from scratch, developing vector-backed RAG search engines, and engineering robust deep learning pipelines in Python and PyTorch.
Building custom Large Language Models from scratch, developing vector-backed RAG search engines, and engineering robust deep learning pipelines in Python and PyTorch.
Engineered a custom GPT architecture entirely from scratch in PyTorch featuring hand-coded multi-head self-attention, token embeddings, and a training loop reducing loss from 10.86 to 1.13.
Built a local Retrieval-Augmented Generation (RAG) pipeline leveraging ChromaDB and Hugging Face sentence-transformers to query project specifications and eliminate model hallucinations.
An end-to-end Python data pipeline and interactive Streamlit web dashboard engineered to clean, process, and visualize cardiovascular health distributions for over 100,000 patients.
Architected a lightweight ranking engine to evaluate 100,000+ records. Engineered a two-phase pipeline using trap filters, maintaining compute limits via sequential file streaming.
Extracted critical business insights identifying content distribution trends. Built comprehensive visual reports mapping historic release velocity and catalog growth metrics.
Engineered a natural language processing pipeline to classify qualitative feedback and quantify user sentiment, translating raw customer reviews into measurable data.