ESPOL (Escuela Superior Politécnica del Litoral) is one of Ecuador’s leading technical universities. I contributed across three parallel roles: ML research, full-stack platform development, and teaching.

ML Engineer — Research

• Built transformer-based segmentation models (Conv-SETR) to automatically identify the subthalamic nucleus and substantia nigra in T1-weighted MRIs, supporting Parkinson's disease treatment planning. Results published in Springer Nature (DICE: 0.81).
• Fine-tuned roBERTa and JobBERTa transformer models for NLP tasks, including sentiment analysis of social media data on urban violence in Guayaquil (ongoing study, 2021–2025).
• Built and integrated a RAG pipeline using LangChain, ChromaDB, and a Django + PostgreSQL backend for an academic research platform.
• Defined and implemented preprocessing pipelines for time series (CPU load) and 3D volumetric images (MRI), using NVIDIA GPU environments on CEDIA infrastructure via Jupyter.
• Created LSTM models to forecast CPU usage in Telconet's datacenter, triggering early warnings against configurable thresholds.
• Tracked training performance with custom Telegram/Webhook callbacks and TensorBoard for metric analysis.

Full Stack Developer — Academic Platform

• Developed the research platform backend with .NET Core 8 and built React data visualizations for research output dashboards.
• Managed containerized deployments with Docker Compose across development and staging environments.
• Freelance contributions to adjacent projects using Astro, TanStack Start, and AWS (EC2, Terraform, Docker).

Teaching Assistant & GenAI Bootcamp Instructor

• Taught GenAI concepts (LLMs, agents, prompt engineering, RAG) to 30+ students as a bootcamp instructor.
• Served as teaching assistant for Object-Oriented Programming: designed exams, graded projects, and ran student office hours.

Tech used

• Python, TensorFlow v2, PyTorch, Scikit-learn
• LangChain, ChromaDB, roBERTa, JobBERTa
• Django, PostgreSQL
• .NET Core 8, React
• Docker Compose, AWS (EC2, Terraform)
• Jupyter, CUDA, Matplotlib, Seaborn, TensorBoard