At RWS Group, I annotate and evaluate multimodal training samples across text, images, video, and audio for LLM training and fine-tuning. I assess model outputs against structured rubrics for factuality, instruction following, and overall quality.
I identify hallucinations, reasoning failures, and edge cases, documenting patterns that inform model iteration and guideline updates. I also perform visual grounding and evaluate audio quality, including pronunciation and prosody.
Previously, as Head of Risk Assessment at Administración Aduanera de Honduras, I led the development of the National Risk Management Manual and Alert Protocol. I also contributed to regional customs risk discussions and applied regulatory risk analysis experience to AI quality assurance.
