Wresty Beralde

Key Skills

Software

Don't disclose

Appen

Top Subject Matter

Text Generation in English

Mapping Business Establishment

Classification of Active Business Establishments

Top Data Types

Text

Image

Top Task Types

Prompt + Response Writing (SFT)

Translation/Localization

Segmentation

Relationship

Classification

Evaluation/Rating

Data Collection

Freelancer Overview

I am an experienced data annotator and evaluator with a strong background in AI training data, specializing in multilingual content, translation evaluation, and sensitive data classification. My work with Appen and global AI projects has involved emotion detection, risk classification, document QA, and the assessment of translation fidelity and toxicity in both English and Tagalog. I have contributed to projects in NLP, social media content relevance, adversarial data annotation, and multimodal AI tasks such as image description and audio evaluation. With over five years of remote workflow management, I am skilled in using Microsoft Office, Google Workspace, and various content management systems, and I consistently apply attention to detail, cultural context awareness, and strict adherence to quality guidelines to ensure the accuracy and reliability of AI datasets.

Labeling Experience

Fake News AI Identifier (Tagalog)

AppenTextRelationshipClassification

Participated in an NDA-protected AI content quality project focused on improving the accuracy and reliability of social media information in the Philippine context. Identified, reviewed, and collected potential misinformation from online platforms, applying analytical judgment, attention to detail, and Tagalog language proficiency. Followed project guidelines and quality standards while contributing consistent work hours to support the development of more trustworthy AI-supported content systems.

Participated in an NDA-protected AI content quality project focused on improving the accuracy and reliability of social media information in the Philippine context. Identified, reviewed, and collected potential misinformation from online platforms, applying analytical judgment, attention to detail, and Tagalog language proficiency. Followed project guidelines and quality standards while contributing consistent work hours to support the development of more trustworthy AI-supported content systems.

2025

Non-stem Evaluation Set Creation

AppenImageRelationshipClassification

Participated in an NDA protected AI data evaluation project focused on building high-quality datasets using image and prompt–response pairs. Contributed to image sourcing and the generation of prompts, responses, and evaluation rubrics to support model assessment and training objectives. Followed strict onboarding guidelines and quality standards to ensure accuracy, relevance, and consistency across all generated data.

Participated in an NDA protected AI data evaluation project focused on building high-quality datasets using image and prompt–response pairs. Contributed to image sourcing and the generation of prompts, responses, and evaluation rubrics to support model assessment and training objectives. Followed strict onboarding guidelines and quality standards to ensure accuracy, relevance, and consistency across all generated data.

2025 - 2025

AI Relevance

AppenTextRelationshipClassification

Participated in an NDA-protected AI content relevance evaluation project focused on improving social media post recommendations by assessing similarity and user-interest alignment between paired posts. Evaluated “Query” and “Nearest Neighbor” content using predefined relevance labels and determined whether users who engaged with the original post would likely find the recommended post interesting. Applied consistent judgment, critical analysis, and strict adherence to quality guidelines to support more engaging and relevant news feed recommendations.

Participated in an NDA-protected AI content relevance evaluation project focused on improving social media post recommendations by assessing similarity and user-interest alignment between paired posts. Evaluated “Query” and “Nearest Neighbor” content using predefined relevance labels and determined whether users who engaged with the original post would likely find the recommended post interesting. Applied consistent judgment, critical analysis, and strict adherence to quality guidelines to support more engaging and relevant news feed recommendations.

2025 - 2025

AI Data Calibration Contributor

AppenImageSegmentation

Contributed to multimodal AI evaluation through image description and audio recording tasks, ensuring high linguistic quality and compliance with calibration standards.

Contributed to multimodal AI evaluation through image description and audio recording tasks, ensuring high linguistic quality and compliance with calibration standards.

2025 - 2025

Toxicity Data Evaluation for Translation

AppenTextTranslation Localization

This task involves evaluating posts in Tagalog-Filipino and its translation in English and evaluating whether a translation preserves the toxicity/sensitivity of the original text. The purpose of this exercise is to identify and classify toxicity equivalence in translations with respect to the original source text. By doing so, we can evaluate the effectiveness of our safety systems in preventing such translations from reaching end-users. NOTE: The content of this task may be toxic in nature and contain things such as profanity, hate-speech, political connotation, negativity, insensitivity to death, gender misrepresentation.

This task involves evaluating posts in Tagalog-Filipino and its translation in English and evaluating whether a translation preserves the toxicity/sensitivity of the original text. The purpose of this exercise is to identify and classify toxicity equivalence in translations with respect to the original source text. By doing so, we can evaluate the effectiveness of our safety systems in preventing such translations from reaching end-users. NOTE: The content of this task may be toxic in nature and contain things such as profanity, hate-speech, political connotation, negativity, insensitivity to death, gender misrepresentation.

2025 - 2025

Show more

Education

Work History