Javier De La Rosa

Key Skills

Software

Label Studio

LightTag

Lionbridge

SuperAnnotate

Telus

Other

Labelbox

Prodigy

Top Subject Matter

No subject matter listed

Top Data Types

Image

Text

Video

Top Task Types

Classification

Entity Ner Classification

Fine Tuning

Prompt Response Writing SFT

Text Summarization

Freelancer Overview

I am an AI Researcher and NLP Specialist with hands-on experience in data labeling, AI training, and model fine-tuning for sentiment analysis, hate speech detection, and content moderation. My expertise lies in leveraging state-of-the-art NLP frameworks such as Hugging Face Transformers, PyTorch, and Gradio to develop impactful AI solutions. I have worked extensively with large-scale datasets, applying advanced annotation techniques, text normalization, and translation pipelines to enhance AI performance. Notably, I fine-tuned HateSpeech-BETO-cased-v2, achieving 84.6% accuracy in detecting nuanced hate speech in Spanish and deployed a real-time Sentiment Analysis Inference Endpoint for social media analytics. My work is instrumental in AI-driven content moderation, customer feedback analysis, and ethical AI development. Passionate about AI fairness and inclusivity, I continuously refine NLP models to better understand human language.

Labeling Experience

Sentiment Analysis for Social Media Content

LabelboxTextClassificationEmotion Recognition

In this project, I worked on labeling social media content and customer feedback to train an AI model for sentiment analysis. The task involved classifying text into sentiment categories (positive, negative, and neutral), taking into account the nuances of human emotion and tone in online communication. By providing highly accurate and consistent annotations, I helped improve the model’s performance in understanding and predicting sentiments expressed in various online platforms. My work ensured that the AI model could better gauge user sentiment, allowing it to respond to customer needs more effectively and enhance brand engagement strategies.

In this project, I worked on labeling social media content and customer feedback to train an AI model for sentiment analysis. The task involved classifying text into sentiment categories (positive, negative, and neutral), taking into account the nuances of human emotion and tone in online communication. By providing highly accurate and consistent annotations, I helped improve the model’s performance in understanding and predicting sentiments expressed in various online platforms. My work ensured that the AI model could better gauge user sentiment, allowing it to respond to customer needs more effectively and enhance brand engagement strategies.

2024

Adds Evaluator

TelusAudioAudio Recording

Project Overview: The project involved assessing and annotating audio advertisements for a global tech company to help train and fine-tune machine learning models used in voice assistants and streaming platforms. The main objective was to evaluate ad relevance, clarity, and user experience within a variety of audio environments (e.g., music playlists, podcasts, smart home interactions). Responsibilities: Evaluated hundreds of audio ads weekly based on pre-defined criteria such as clarity, sentiment, relevance to the user’s interest, and compliance with platform guidelines. Annotated audio segments for background noise, voice clarity, speaker emotion, and ad length compliance. Identified and tagged ad content for sensitive or inappropriate language, brand mentions, call-to-action effectiveness, and region-specific targeting. Provided detailed feedback on ads that failed quality thresholds, helping clients improve ad scripts and delivery.

Project Overview: The project involved assessing and annotating audio advertisements for a global tech company to help train and fine-tune machine learning models used in voice assistants and streaming platforms. The main objective was to evaluate ad relevance, clarity, and user experience within a variety of audio environments (e.g., music playlists, podcasts, smart home interactions). Responsibilities: Evaluated hundreds of audio ads weekly based on pre-defined criteria such as clarity, sentiment, relevance to the user’s interest, and compliance with platform guidelines. Annotated audio segments for background noise, voice clarity, speaker emotion, and ad length compliance. Identified and tagged ad content for sensitive or inappropriate language, brand mentions, call-to-action effectiveness, and region-specific targeting. Provided detailed feedback on ads that failed quality thresholds, helping clients improve ad scripts and delivery.

2023 - 2023

Internet Safety Evaluation for Google AI

ProdigyTextClassificationEmotion Recognition

As part of a large-scale initiative to improve AI’s ability to detect harmful online content, I conducted data labeling for Google's safety evaluation models. This project required me to classify web content into different categories: safe, unsafe, or ambiguous. I ensured that my annotations were precise and aligned with the safety standards, contributing to the AI model's ability to automatically flag inappropriate content. The project enhanced the overall safety of online platforms and helped refine AI algorithms for more efficient content moderation.

As part of a large-scale initiative to improve AI’s ability to detect harmful online content, I conducted data labeling for Google's safety evaluation models. This project required me to classify web content into different categories: safe, unsafe, or ambiguous. I ensured that my annotations were precise and aligned with the safety standards, contributing to the AI model's ability to automatically flag inappropriate content. The project enhanced the overall safety of online platforms and helped refine AI algorithms for more efficient content moderation.

2019 - 2020

Education

Work History