Quantization, rotations, implementable primitives and algorithm-hardware co-design for more efficient AI models.

Efficient and hardware-aware AI

Methods, manifests and benchmarks for detecting temporal shortcuts and evaluating medical models with more rigor.

Anti-leakage evaluation

Modular architectures for reproducible, traceable medical research prepared for multimodal validation.

Clinical-Core

An editorial lab for turning reading, research and technical practice into essays, books and learning resources.

FP32 and technical books

I use demanding medical problems as a testbed for hardware-aware optimization, rigorous evaluation and tools that connect machine learning, quantization, RTL and verifiable systems.

Research lines by Raul Pacheco Rodriguez in efficient AI, anti-leakage evaluation and algorithm-hardware co-design.

Efficient, reproducible and deployable AI.

Diseño protocolos y herramientas para detectar cuándo una IA médica está aprendiendo del futuro, de sesgos del pipeline o de atajos metodológicos.

Evaluación confiable y anti-leakage en IA médica

Construyo pipelines médicos multimodales que pueden auditarse, intercambiar módulos y evaluar cada modalidad sin convertir todo en una caja negra.

Arquitecturas multimodales modulares para pronostico medico 

Investigo cómo comprimir modelos médicos mediante rotaciones ortogonales para que funcionen en baja precisión sin perder utilidad clínica downstream.

Efficient, reproducible and deployable AI.

Evaluación confiable y anti-leakage en IA médica

Arquitecturas multimodales modulares para pronostico medico

Cuantización rotacional y ML eficiente hardware-aware