About Work Focus Links

Raúl Alejandro Pérez Saucedo

I’m interested in building AI systems for policy, law, and climate.

I work on applying large language models to real-world problems, including legal analysis and policy research, where I focus on transforming complex, multilingual texts into structured data. My goal is to make these systems accessible, reliable, and useful beyond expert users.

```
Portrait of Raúl Pérez Saucedo
Background
B.S. in Data Science and Mathematics
Tecnológico de Monterrey (ITESM)

About

I currently collaborate with The Carter Center, where I build LLM-based workflows to analyze legal texts on freedom of expression and association. My work sits at the intersection of AI and policy with a particular interest in Latin America.

During my undergraduate studies, I worked across both research and industry. I received the Mitacs Globalink Research Scholarship to work on AI pipelines for extracting and structuring NGO-regulation data at Concordia University. At CEMEX, I built SQL-based data systems and dashboards for operational metrics.

What I care about
Making AI systems accessible to everyday users, not just technical experts.
Long-term direction
Building fair and sustainable AI systems for policy and climate applications.

Projects

LLM Pipeline for Legal Coding

I designed and evaluated a two-stage LLM pipeline to support the coding of legal texts on freedom of expression. The goal was not full automation, but building a system that meaningfully reduces the burden on human researchers.

  • Processed a global corpus of 160+ laws across 60+ countries
  • Built a recall–verification pipeline combining GPT models, retrieval, and structured outputs
  • Showed tradeoffs between model performance, cost, and reliability in real-world workflows
Report ↗ Code ↗

Comparative NGO Regulations with LLMs

I developed and tested an LLM-based workflow for analyzing NGO-related laws across countries, focused on how prompting strategies and model design affect reliability in comparative legal research.

  • Evaluated performance across 10 countries and 3 languages
  • Compared prompting strategies, showing simple rule-based prompts outperform complex reasoning
  • Developed a prompt development framework for legal coding tasks
  • Designed workflows enabling scalable, cross-national legal data collection
Paper ↗ Code ↗

Current focus

Right now, I’m most interested in applied LLM research: retrieval-augmented generation, agentic workflows, and evaluation methods that help these systems become more reliable and helpful for people.

Interests

Applied LLM systems Multilingual RAG AI accessibility Public policy in Latin America Climate
```