Isaiah Villalobos

Developing data pipelines and retrieval systems for an AI support platform that transforms unstructured archives into searchable, citation-backed knowledge.

• Built ETL pipelines to scrape, parse, and clean 20+ years of AMBER mailing list archives and tutorials using BeautifulSoup, regex, and Python, producing 42,000+ structured documents for downstream vector indexing.
• Developed automated data ingestion workflows using Apache Airflow to schedule daily scraping, HTML to JSON conversion, metadata extraction, and ChromaDB indexing tasks.
• Implemented a RAG pipeline using ChromaDB vector search and FAISS-based PDF retrieval to generate responses with automated citation links, served via a Llama 3.1 inference server on Linux HPC infrastructure.

Building and maintaining production web systems with a focus on accessibility, authentication, backend APIs, and end-to-end delivery.

• Architected and shipped RESTful CRUD APIs for BOD document management and announcement banners using Supabase and Node.js, with the banner API serving all pages across 100,000+ annual visits.
• Built a student contact form API with rate limiting, input validation, and JotForm integration using Next.js API routes and Node.js, solving a gap where students had no reliable way to submit concerns to USU administration.
• Implemented secure SSO authentication using NextAuth and Microsoft Azure AD with role-based access control for 6 administrative users, following service-oriented architecture patterns across internal back-office systems.
• Fully designed and shipped the Wingspan program page from scratch and led major rebuilds of FSL and U-Krew, owning the full SDLC from requirements gathering through solo deployment on Vercel.
• Maintained zero WAVE accessibility errors and strong Google Lighthouse scores across all production pages through responsive design, image optimization, and AI search indexing following Agile development workflows.