Juan Osuna

Key Skills

Software

Labelbox

Prodigy

Label Studio

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Image

Text

Top Task Types

Audio Recording

Classification

Computer Programming Coding

Evaluation Rating

Translation Localization

Freelancer Overview

Bilingual AI linguist and NLP specialist with native fluency in Mexican Spanish and advanced English (C1), holding an M.Sc. in Artificial Intelligence from Monterrey Tech. I specialize in culturally grounded AI data annotation, prompt engineering, and linguistic localization for NLP systems. At Softek and CIITEC, I annotated 15,000+ high-precision Spanish-Mexican samples using Label Studio and Prodigy, maintaining 98.7% accuracy while reducing team inconsistencies by 35%. I designed culturally adapted prompt sets that boosted model performance by 22–27% for regional audiences and implemented QA protocols that cut annotation errors by 40%. My academic research -“Cultural Adaptation of Prompt Engineering for AI Systems in Mexican Spanish Contexts”- directly informs my professional approach, ensuring linguistic nuance, dialect awareness, and cultural relevance in every dataset. I collaborate closely with ML engineers to refine guidelines, accelerate workflows (15% faster turnaround), and uphold strict ethical, privacy, and bias-aware standards, especially when handling sensitive content. Detail-oriented, deadline-driven, and fluent in technical communication across languages, I deliver training data that bridges cultural authenticity with AI performance.

Labeling Experience

AI Training Data Annotation & Prompt Engineering for Multilingual LLMs

ProdigyTextEntity Ner ClassificationClassification

I am currently contributing to a high-volume NLP project supporting the development of multilingual large language models (LLMs) for global clients. To date, I have annotated over 5,000 linguistic samples monthly in Mexican Spanish, consistently maintaining a 98.7% precision rate. I translate and generate culturally adapted prompt sets leveraging regional expressions and sociolinguistic context, resulting in a 22% measured improvement in model accuracy for Mexican Spanish users. I also designed standardized annotation protocols that reduced team inconsistencies by 35%, and collaborate directly with ML engineers to refine guidelines, accelerating task completion by 15% while preserving quality. All work follows strict ethical, privacy, and bias-aware standards, including responsible handling of sensitive content. This ongoing role demonstrates my ability to deliver scalable, culturally grounded training data that enhances AI performance in real-world applications.

I am currently contributing to a high-volume NLP project supporting the development of multilingual large language models (LLMs) for global clients. To date, I have annotated over 5,000 linguistic samples monthly in Mexican Spanish, consistently maintaining a 98.7% precision rate. I translate and generate culturally adapted prompt sets leveraging regional expressions and sociolinguistic context, resulting in a 22% measured improvement in model accuracy for Mexican Spanish users. I also designed standardized annotation protocols that reduced team inconsistencies by 35%, and collaborate directly with ML engineers to refine guidelines, accelerating task completion by 15% while preserving quality. All work follows strict ethical, privacy, and bias-aware standards, including responsible handling of sensitive content. This ongoing role demonstrates my ability to deliver scalable, culturally grounded training data that enhances AI performance in real-world applications.

2024

Research-Driven NLP Data Annotation & Dialect-Specific Prompt Development

Label StudioTextClassificationText Generation

I supported an academic research initiative focused on improving AI performance in Mexican Spanish contexts. I annotated and curated over 10,000 Spanish-Mexican language samples for training NLP models, ensuring high fidelity in dialect-specific expressions and cultural references. I designed and tested high-quality prompt sets that increased AI response accuracy for local users by 27%. Additionally, I implemented rigorous QA protocols that reduced annotation errors by 40%, while documenting regional linguistic variations to inform model adaptation strategies. My work directly contributed to the thesis "Cultural Adaptation of Prompt Engineering for AI Systems in Mexican Spanish Contexts," aligning technical execution with academic rigor. This project demonstrates my expertise in combining deep linguistic understanding with practical NLP workflows for research-grade AI systems.

I supported an academic research initiative focused on improving AI performance in Mexican Spanish contexts. I annotated and curated over 10,000 Spanish-Mexican language samples for training NLP models, ensuring high fidelity in dialect-specific expressions and cultural references. I designed and tested high-quality prompt sets that increased AI response accuracy for local users by 27%. Additionally, I implemented rigorous QA protocols that reduced annotation errors by 40%, while documenting regional linguistic variations to inform model adaptation strategies. My work directly contributed to the thesis "Cultural Adaptation of Prompt Engineering for AI Systems in Mexican Spanish Contexts," aligning technical execution with academic rigor. This project demonstrates my expertise in combining deep linguistic understanding with practical NLP workflows for research-grade AI systems.

2023 - 2024

Education

Work History