Architected and deployed an event-driven n8n workflow triggered via Google Drive API to automate pre-screening and audit text against ambiguous project-specific compliance rules in Spanish.

Architected an end-to-end LangChain RAG-LLM application to match 150+ researchers across disciplines. Managed local inference on an NVIDIA H100 using Ollama (Llama 3, Mixtral) and OpenAlex JSON metadata.

Developed an ML model processing a 20GB dataset of EEG signals and Spotify metadata, achieving a 20% improvement in F1-score over the baseline. Published in IEEE ICASSP proceedings.

Contributed to deep learning models identifying clinical artifacts (air bubbles, missing lids) in Whole Slide Images (WSIs) using PyTorch, OpenCV, and OpenSlide for gigapixel medical imaging triage, achieving 94% accuracy.

Architected a full-stack graph engine transforming 80+ GB Wikipedia MySQL dumps into an interactive WebGL visualization. Built a high-throughput Python/Bash ETL pipeline for PostgreSQL ingestion and sub-second API payloads to a Sigma.js frontend.