Jonathan Ferreiras

Key Skills

Software

Appen

CrowdSource

Google Cloud Vertex AI

Labelbox

Label Studio

OneForma

Remotasks

Scale AI

Toloka

Other

Internal/Proprietary Tooling

Top Subject Matter

Prompt generation

Prompt evaluation

Response evaluation and generation

Top Data Types

Audio

Image

Text

Top Task Types

Segmentation

Text Generation

Translation Localization

Freelancer Overview

Highly motivated Industrial Engineer with over two years of dedicated experience in Generative AI projects, transitioning expertise in quality, data analytics, and business intelligence to the cutting-edge field of AI training. Proven ability to deliver high-quality annotated datasets through meticulous data annotation, rigorous response validation, and precise polygon segmentation. Adept at crafting effective prompts for Large Language Models (LLMs), optimizing their performance, and ensuring accurate transcription and Text-to-Speech (TTS) outputs. Passionate about contributing to the advancement of AI, I bring a strong analytical mindset and a commitment to excellence to every project.

Labeling Experience

Generative AI Annotator / TTS / Transcriber

CrowdsourceAudioSegmentationText Generation

● Label, categorize, and tag data. ● Classify content and text based on different projects guidelines. ● Identify and label named entities. ● Mark instances of biased or toxic language. ● Rank multiple LLM responses based on quality. ● Label and adjust audio parameters. ● Improve the accuracy and reliability of LLM outputs. ● Make sure generated speech sounds accurate natural, aligned to the project’s guidelines. ● Review and correct audio outputs produced by the TTS systems. ● Listen and transcribe.

● Label, categorize, and tag data. ● Classify content and text based on different projects guidelines. ● Identify and label named entities. ● Mark instances of biased or toxic language. ● Rank multiple LLM responses based on quality. ● Label and adjust audio parameters. ● Improve the accuracy and reliability of LLM outputs. ● Make sure generated speech sounds accurate natural, aligned to the project’s guidelines. ● Review and correct audio outputs produced by the TTS systems. ● Listen and transcribe.

2024 - 2024

Data Annotator

CrowdsourceImageBounding BoxPolygon

● Label, categorize, and tag data. ● Classify content and text based on different projects guidelines. ● Identify and label named entities. ● Mark instances of biased or toxic language. ● Rank multiple LLM responses based on quality. ● Improve the accuracy and reliability of LLM outputs

● Label, categorize, and tag data. ● Classify content and text based on different projects guidelines. ● Identify and label named entities. ● Mark instances of biased or toxic language. ● Rank multiple LLM responses based on quality. ● Improve the accuracy and reliability of LLM outputs

2024

AI Trainer

Internal Proprietary ToolingTextEvaluation RatingPrompt Response Writing SFT

1. Ensuring dataset quality by cleaning, structuring and labelling data. 2. Annotate data adding tags and label according to context. 3. Ensure the model behave as intended, addressing bias, inaccuracies, and undesired output

1. Ensuring dataset quality by cleaning, structuring and labelling data. 2. Annotate data adding tags and label according to context. 3. Ensure the model behave as intended, addressing bias, inaccuracies, and undesired output

2024 - 2024

Prompt Evaluator and Annotator

Label StudioTextEvaluation Rating

1. Analyze queries and understand the user’s intent and ensure the context is valid. 2. Classify the user's intent and identify the query type and entity type based on customer intent and entity. 3. Evaluate two responses and assess each response for harmlessness, relevance, helpfulness, consistency, clarity, objectivity, and tone. 4. Rank the answers and choose the better answer based on the quality criteria, provide reasons for the preference, and add any relevant comments. 5. Leave comments if necessary.

1. Analyze queries and understand the user’s intent and ensure the context is valid. 2. Classify the user's intent and identify the query type and entity type based on customer intent and entity. 3. Evaluate two responses and assess each response for harmlessness, relevance, helpfulness, consistency, clarity, objectivity, and tone. 4. Rank the answers and choose the better answer based on the quality criteria, provide reasons for the preference, and add any relevant comments. 5. Leave comments if necessary.

2023 - 2023

Prompt and Response Evaluator (web search)

CrowdsourceTextEvaluation RatingPrompt Response Writing SFT

1. Analyze queries and understand the user’s intent and ensure the context is valid. 2. Classify the user's intent and identify the query type and entity type based on customer intent and entity. 3. Evaluate two responses and assess each response for harmlessness, relevance, helpfulness, consistency, clarity, objectivity, and tone. 4. Rank the answers and choose the better answer based on the quality criteria, provide reasons for the preference, and add any relevant comments. 5. Leave comments if necessary.

1. Analyze queries and understand the user’s intent and ensure the context is valid. 2. Classify the user's intent and identify the query type and entity type based on customer intent and entity. 3. Evaluate two responses and assess each response for harmlessness, relevance, helpfulness, consistency, clarity, objectivity, and tone. 4. Rank the answers and choose the better answer based on the quality criteria, provide reasons for the preference, and add any relevant comments. 5. Leave comments if necessary.

2023 - 2023

Education

Work History