Systems and Computer Science Engineer at the National University of Colombia. I build agentic AI systems, computer vision pipelines, and ML products that reach production, from industrial defect detection to multi-agent marketing assistants.
Currently working at ReshapeX, building AI agents for industrial companies.
🇨🇴 Based in Colombia, home to the world's largest variety of butterflies and birds, and the second most biodiverse country on Earth.
- Ecosistema de Datos (MinTIC 2025): Team project building a fully reproducible pipeline to process and analyze 330k crime records across Santander. Forecasting (Prophet, SARIMA), spatial hotspot detection (DBSCAN, KDE), and XGBoost risk classification, delivering dashboards and actionable insights for public safety.
- AI Data Challenge (Source Meridian 2025): End-to-end machine learning pipeline for real-world datasets. Team achieved 2nd place overall with a classification system using XGBoost and automated data processing.
- Training Camp Medellín, Competitive Programming (EAFIT 2024): Intensive algorithmic problem-solving camp, representing Universidad Nacional de Colombia.
B.S. in Systems Engineering and Informatics · Universidad Nacional de Colombia (2027) Coursework: Data Structures, Algorithms, Machine Learning Fundamentals, Database Systems, and more.
Certifications: Deep Learning with PyTorch (Coursera), Neural Networks and Deep Learning (DeepLearning.AI), AWS Academy Cloud Foundations. Full list on LinkedIn.
Full write-ups and the complete list live on the portfolio site.
|
Production hospitality platform: concurrency-safe booking, role-based access control, and a full unit/integration/E2E/load test pyramid under real traffic bounds.
|
330k crime records turned into forecasts, spatial hotspots, and risk classification, with a RAG layer over the underlying documents for public safety intelligence.
|
|
Computer vision pipeline on a SICK industrial camera catching metal-sheet defects on a live frying-pan production line, from acquisition through real-time inference.
|
Specialized agents for analytics, content, and project management coordinating through shared memory and dynamic routing, with versioned prompts for traceability.
|
|
Transfer-learned CNNs judged on clinical metrics, sensitivity and specificity over raw accuracy, designed as an assisted screening tool for low-resource settings.
|
RAG-based data extraction tool pulling explicit terms from Commercial Real Estate Lease Agreements. Freelance collaboration.
|
Agents & RAG
ML & Computer Vision
Backend & Web
Data & Infrastructure
Currently Learning
🇨🇴 Open to collaborations on applied AI and multi-agent systems.