Andreia Silva

Key Skills

Software

Mindrift

Other

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Classification

Evaluation Rating

Text Generation

Freelancer Overview

I am an AI Data Annotation and Model Evaluation specialist with hands-on experience in LLM prompt and response evaluation, text generation, classification, and safety review. I work across multiple platforms including Uber AI Solutions, Mindrift/Tendem, and OpenTrain AI, where I evaluate AI outputs for accuracy, clarity, guideline compliance, and safety alignment. My professional background in operations, incident investigation, fraud analysis, and escalation workflows gives me strong analytical judgment and attention to detail—skills that directly support high-quality AI training data. I am experienced in identifying patterns, detecting anomalies, reviewing reasoning quality, and applying structured decision frameworks to complex tasks. I contribute high-quality training data by creating prompts, generating ideal model responses, rewriting text for clarity or tone, and assessing multi-step reasoning. I am also proficient in safety classification, appropriateness review, and content quality evaluation. I follow detailed guidelines carefully, document findings clearly, and consistently produce reliable, accurate annotations that help improve AI model performance and behaviour.

Labeling Experience

Text Generation & Model Training Data Creation

OtherTextQuestion AnsweringText Generation

Created high-quality training data to support the development of large language models. Tasks included writing diverse user prompts, generating ideal assistant responses, rewriting text for clarity and tone, and producing structured dialogues across multiple topics. Ensured linguistic accuracy, naturalness, and alignment with task-specific guidelines. Contributed creative, well-structured, and instruction-following text samples to enhance AI reasoning, fluency, and user-experience performance.

Created high-quality training data to support the development of large language models. Tasks included writing diverse user prompts, generating ideal assistant responses, rewriting text for clarity and tone, and producing structured dialogues across multiple topics. Ensured linguistic accuracy, naturalness, and alignment with task-specific guidelines. Contributed creative, well-structured, and instruction-following text samples to enhance AI reasoning, fluency, and user-experience performance.

2025

LLM Prompt & Response Evaluation

MindriftTextClassificationText Generation

I evaluated AI-generated responses for clarity, accuracy, safety, and adherence to user instructions. Tasks included rating model outputs, identifying errors, improving responses, and ensuring alignment with detailed quality guidelines. I worked with multiple prompt types such as creative writing, classification, summarization, and multi-step reasoning. Responsibilities also included assessing safety concerns, reviewing policy compliance, and providing structured feedback to improve LLM behavior and reliability.

I evaluated AI-generated responses for clarity, accuracy, safety, and adherence to user instructions. Tasks included rating model outputs, identifying errors, improving responses, and ensuring alignment with detailed quality guidelines. I worked with multiple prompt types such as creative writing, classification, summarization, and multi-step reasoning. Responsibilities also included assessing safety concerns, reviewing policy compliance, and providing structured feedback to improve LLM behavior and reliability.

2025 - 2025

AI Safety Rating & Content Quality Review

MindriftTextClassificationEvaluation Rating

Reviewed text content for safety, policy compliance, and guideline adherence. Classified user prompts and AI responses into allowed and disallowed categories, identified harmful or sensitive content, and rated severity levels. Performed quality assessments of reasoning, tone, and factual correctness. Applied detailed policy frameworks to ensure alignment with safety, bias, and appropriateness standards. Provided structured feedback and rationale to support safer, more reliable AI models.

Reviewed text content for safety, policy compliance, and guideline adherence. Classified user prompts and AI responses into allowed and disallowed categories, identified harmful or sensitive content, and rated severity levels. Performed quality assessments of reasoning, tone, and factual correctness. Applied detailed policy frameworks to ensure alignment with safety, bias, and appropriateness standards. Provided structured feedback and rationale to support safer, more reliable AI models.

2024 - 2025

Education

Work History