Nancy U. Morales

Key Skills

Software

Remotasks

Other

Scale AI

Appen

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Text

Video

Top Task Types

Data Collection

Emotion Recognition

Evaluation Rating

Prompt Response Writing SFT

Translation Localization

Freelancer Overview

I’m a detail-oriented freelancer with proven experience in AI data labeling, annotation, and evaluation. I’ve contributed to multiple international projects through Appen and OneForma, performing diverse tasks such as transcription, translation evaluation, speaker diarization, ad quality review, and LLM reasoning assessment. What sets me apart is my strong analytical mindset, accuracy under pressure, and ability to quickly adapt to different project requirements. I have a solid background in customer and data analysis, quality assurance, and process optimization, which allows me to maintain consistency and deliver high-quality results. My focus on precision, reliability, and clear communication makes me a valuable collaborator for any AI data or evaluation project.

Labeling Experience

Text Moderation

OtherTextClassificationQuestion Answering

Accuracy 90% The goal of the project is to identify and classify inappropriate language within a gaming environment. The task involves labeling and categorizing text data so that the AI system can be trained to automatically detect such content in the future. The main objective is to prevent and filter out text that violates content policies related to hate speech, sexual content, sexual harassment, threats, wishing harm, and violent speech. Cinder is the name of the platform used for this tasks.

Accuracy 90% The goal of the project is to identify and classify inappropriate language within a gaming environment. The task involves labeling and categorizing text data so that the AI system can be trained to automatically detect such content in the future. The main objective is to prevent and filter out text that violates content policies related to hate speech, sexual content, sexual harassment, threats, wishing harm, and violent speech. Cinder is the name of the platform used for this tasks.

2025

Translation Evaluation

AppenAudioQuestion AnsweringEmotion Recognition

Accuracy 90% Task consisted on evaluate a translation generated by a AI model, evaluate the Faithfulness, Expressiveness (Speech Rhythm,Speaking Rate,Loudness,Pitch,Prosody)and Overall Quality. Consisted in Evaluating how well each translation captures both the meaning and speaking style of the original audio. The primary focus is how the translations sound compared to the source audio. Assessments contribute to the development of more effectiveand natural-sounding speech to speech translation systems.

Accuracy 90% Task consisted on evaluate a translation generated by a AI model, evaluate the Faithfulness, Expressiveness (Speech Rhythm,Speaking Rate,Loudness,Pitch,Prosody)and Overall Quality. Consisted in Evaluating how well each translation captures both the meaning and speaking style of the original audio. The primary focus is how the translations sound compared to the source audio. Assessments contribute to the development of more effectiveand natural-sounding speech to speech translation systems.

2025

Prompt Responses Evaluation

AppenTextQuestion AnsweringPrompt Response Writing SFT

Reviewing prompts and their responses, evaluating the quality of responses that a bot would produce in response to various prompts. The evaluation helps ensure high-quality data will be used to generate synthetic evaluation data for Large Language Models (LLMs). Rating categories: Harm Correctness Relevance

Reviewing prompts and their responses, evaluating the quality of responses that a bot would produce in response to various prompts. The evaluation helps ensure high-quality data will be used to generate synthetic evaluation data for Large Language Models (LLMs). Rating categories: Harm Correctness Relevance

2025 - 2025

Reels Translation and Auto Dubbing

AppenAudioQuestion AnsweringText Generation

Accuracy 85% An audio file contanied an excerpt of speech. A transcription came along as well. Annotators were required to review and correct the transcription so that it matches the excerpt of speech, with exceptions for technical issues. Add Audio labels Answer a Yes/No question regarding the quality and nature of the audio clip Check a box if there are background noises or if speech is cut off in the audio clip

Accuracy 85% An audio file contanied an excerpt of speech. A transcription came along as well. Annotators were required to review and correct the transcription so that it matches the excerpt of speech, with exceptions for technical issues. Add Audio labels Answer a Yes/No question regarding the quality and nature of the audio clip Check a box if there are background noises or if speech is cut off in the audio clip

2025 - 2025

Speaker Diarization

AppenAudioSegmentation

The project focuses on speaker diarization, which involves identifying and distinguishing individual speakers within an audio recording. The main goal is to segment the audio into parts corresponding to each speaker, allowing the AI system to recognize who spoke when. Tasks include carefully listening to audio files, marking changes in speakers, aligning transcriptions with the correct speaker segments, placing and verifying the accuracy of timestamps. This process requires strong attention to detail, active listening skills, and consistency to ensure precise labeling. By providing accurate speaker segmentation and labeling, the project helps improve automatic speech recognition systems and conversational AI models, enabling them to handle multi-speaker audio more effectively and naturally.

The project focuses on speaker diarization, which involves identifying and distinguishing individual speakers within an audio recording. The main goal is to segment the audio into parts corresponding to each speaker, allowing the AI system to recognize who spoke when. Tasks include carefully listening to audio files, marking changes in speakers, aligning transcriptions with the correct speaker segments, placing and verifying the accuracy of timestamps. This process requires strong attention to detail, active listening skills, and consistency to ensure precise labeling. By providing accurate speaker segmentation and labeling, the project helps improve automatic speech recognition systems and conversational AI models, enabling them to handle multi-speaker audio more effectively and naturally.

2025 - 2025

Show more

Education

Work History