LLM Semantic QA · Cultural & Factuality Evaluation · Language Model Behavior Analysis

I work on qualitative evaluation of large language models, focusing on failure modes that are often missed by standard benchmarks: confidence miscalibration, semantic over-interpretation, cultural misrecognition, and pragmatic errors.

My background combines:

• multilingual linguistic QA (EN / ES)
• editorial accuracy and content evaluation
• applied analysis of how LLMs interpret, infer, and sometimes overreach

Rather than optimizing prompts for output quality alone, I document where and why models fail to understand language as humans do (especially in culturally localized, implicit, or ambiguous contexts).

Some essays, reflections, and case studies live on my Substack: 👉 https://open.substack.com/pub/alejandroremeseiro

• Semantic and pragmatic QA for LLMs
• Model confidence calibration (when models should ask instead of infer)
• Cultural reference misrecognition (implicit humor, local context)
• Factuality vs. coherence analysis
• Spanish (Spain)–specific AI evaluation

• hallucination-cases Documented examples of structural and epistemic hallucinations in LLM outputs

• prompt-eval-cases
  Real-world prompt cases highlighting semantic and cultural failure modes (EN/ES)

• factuality-bias-review
  Annotated examples of hallucinations, inconsistencies, and bias patterns

• BA in History, University of Alcalá (2006)
• Darmasiswa scholar, Indonesia (Surabaya, 2009–2010)
• 14+ years working with international clients in linguistic quality and content review

Open to remote collaboration on:

• LLM evaluation and QA
• Semantic & cultural error analysis
• Linguistic and bilingual (EN/ES) model evaluation

Email: alejandro.remeseiro(at)gmail.com

LinkedIn: https://www.linkedin.com/in/alejandro-remeseiro-fern%C3%A1ndez-44a02427/

Upwork profile: https://www.upwork.com/freelancers/~015bb79bd0df3c5e7f