Carlos Andrés Durán Paredes

I'm a Master of Engineering student at Universidad de Antioquia (UdeA) working on quantum machine learning: variational classifiers, hybrid quantum-classical models, and experiments on real NISQ hardware through IBM Quantum. I graduated in Physics Engineering from Universidad del Cauca.

My work sits between papers and packages. When I publish a result, I try to ship the code as something other people can install, rerun, and check, with honest baselines and leakage-free evaluation instead of headline "quantum advantage" claims.

Before quantum, I worked on deep learning for urban acoustics (the Sound Maps project), led the IEEE AESS student chapter at Unicauca, and took part in the university's aerospace research seedbed.

qiskit-data-reuploading

A scikit-learn-compatible data re-uploading quantum classifier (Pérez-Salinas et al., 2020) built on Qiskit 2.x V2 primitives. It is listed in the official Qiskit Ecosystem as a paper artifact and runs on simulators or IBM Quantum hardware with the same fit / predict API.

pip install qiskit-data-reuploading

TLM-UAV Quantum Anomaly Detection  
Hybrid XGBoost + data re-uploading detector for cyber-physical attacks on drones, evaluated with a group-aware temporal split and a three-level proxy-feature audit. The hybrid is the only model whose F1 improves once contextual proxies are removed, and it records the lowest false-alarm rate under proxy-free evaluation. Fully reproducible.

GCM Strategy: Quantum Battle of the Sexes on NISQ hardware  
Classical vs. quantum (EWL scheme) Battle of the Sexes across 31 entanglement values, run on ideal simulation, noisy simulation, and real IBM QPUs. A guided circuit-mapping strategy picks qubit pairs to reduce error on hardware.

UrbanSonic-CNNs  
CNN classifier for the Urbanphony urban soundscape taxonomy on spectrograms from historic cities. Published in Ingeniería e Investigación and at IntelliSys 2025.

Operational insights chatbot: natural-language KPI queries over logistics data. The LLM only parses intent and writes the narrative; every number comes from pandas, so answers can't hallucinate figures.

Competitive intelligence pipeline: price, fee, ETA, and promotion monitoring across three delivery platforms in Mexico, with a Streamlit dashboard and an HTML executive report.

• Durán Paredes, C. A. et al. Quantum Machine Learning for Cyber-Physical Anomaly Detection in Unmanned Aerial Vehicles. arXiv:2605.19233, 2026.
• Díaz Agreda, G. D., … Durán Paredes, C. A. et al. Experimental Implementation of the Quantum Volunteer's Dilemma on NISQ Hardware. arXiv:2605.30676, 2026.
• Cajas Ordóñez, S. A., … Durán, C. A. et al. Uncertainty Makes It Stable: Curiosity-Driven Quantized Mixture-of-Experts. arXiv:2511.11743, 2025.
• Cajas Ordóñez, S. A., … Durán, C. A. et al. Embedding-Aware Quantum-Classical SVMs for Scalable Quantum Machine Learning. arXiv:2508.00024, 2025.
• Díaz Agreda, G. D., Durán Paredes, C. A. et al. Quantum Battle of the Sexes with Error Mitigation on NISQ Hardware. IEEE CHILECON, 2025.
• Durán Paredes, C. A. et al. Urbanphony-3-CNN. Ingeniería e Investigación 45(2), 2025.
• Durán, C. et al. A CNN-Based Approach for Classifying Urban Soundscape Taxonomy in Historic Cities. IntelliSys, 2025.

Full list on Google Scholar.