OmarLópez
I specialize in machine learning, deep learning, and AI-powered solutions — from privacy-preserving synthetic data on HPC clusters to production RAG systems and LLM agents serving 100k+ clients.
RAGMLOpsHPC
PyTorchAgents
Areas of Deep Work
Secured Synthetic Mammogram Generation
Contributed to the European SECURED Project advancing privacy-preserving synthetic data generation for medical imaging. Designed and implemented scalable GANs and Diffusion Models on HPC environments (Marenostrum 5), incorporating differential privacy to comply with EU regulations.
Read Paper →SVM Kernel Development for Omic Data
Designed novel Jaccard-Tanimoto-based kernels for SVMs tailored to high-dimensional sparse omic datasets. Proved generalization to real-valued feature spaces, demonstrating superior accuracy vs. RBF kernels in cancer subtype classification and gene expression analysis.
Read Paper →Peer-Reviewed Work
Peer Reviewed
SynthVal: A Framework for Validating Synthetic Medical Images
A Python framework for validating synthetic medical image quality through statistical comparisons in deep feature space. Uses transformer-based models to extract semantic embeddings and computes Fréchet Distance, Wasserstein Distance, and KL Divergence between real and synthetic distributions. Evaluated on the CSAW-CC mammography dataset with images from the Barcelona Supercomputing Center.
↗ View on ResearchGateExperience & Education
Design, development, and deployment of Generative AI solutions for multiple enterprise clients. Integrating LLM-based capabilities into data platforms to modernize analytics and enhance decision-making.
Architected a real-time AI agent automating contract sales for 100,000+ clients. Engineered an Agent-to-Agent (A2A) communication protocol and automated pipeline ingesting 100,000+ contract pages monthly.
Generation and security of diffusion models in HPC environments. MLOps solutions for serving ML applications. European SECURED project on privacy-preserving medical imaging.
University and master-level courses in programming, statistics, and machine learning. 100+ students across Python, Java, R, ML, and data science.
Led an OCR team on a complex information extraction project, delivering foundations for ML-based document classification tools.