Joshua Pereira

Key Skills

Software

Appen

AWS SageMaker

Clickworker

Data Annotation Tech

Dataloop

Deep Systems

Diffgram

Top Subject Matter

No subject matter listed

Top Data Types

Text

Computer Code Programming

Document

Top Label Types

Classification

Computer Programming Coding

Data Collection

Object Detection

Polygon

Evaluation Rating

Transcription

Prompt Response Writing SFT

Fine Tuning

Text Summarization

Text Generation

Question Answering

Freelancer Overview

I am a machine learning engineer with over two years of hands-on experience building and optimizing data-driven solutions, with a strong focus on data labeling, annotation, and AI training data pipelines. My expertise spans Python, Pandas, NumPy, Sci-learn, and TensorFlow, which I have applied to real-world projects such as medical insurance prediction, breast cancer classification, and stock price forecasting, all of which required meticulous data preprocessing and annotation for model accuracy. I have also worked on computer vision tasks, including ProofPulse, an AI-powered deep learning tool for detecting AI-generated images, where high-quality labeled data was critical for model performance. My background in NLP, predictive modeling, and end-to-end deployment using tools like Streamlit, SQL, and GitHub enables me to transform raw datasets into production-ready machine learning systems. I am passionate about ensuring data quality and integrity throughout the AI lifecycle, and I am adept at using both manual and automated tools for data annotation and validation.

Labeling Experience

Proofpulse

AppenImageClassificationObject Detection

Scope of the Project This project focuses on the design and development of a web-based hybrid AI-image detection system (ProofPulse) capable of distinguishing between authentic and AI-generated images. The system integrates custom-trained deep learning models with external detection services to cross-validate results and present confidence-based classifications through an interactive user interface. The scope covers dataset preparation, model training, system integration, evaluation, and deployment of a user-accessible platform, but does not include real-time social media scraping, video-based deepfake detection, or large-scale cloud production deployment. Specific Data Labelling Tasks To ensure accurate supervised learning, structured data labelling activities were carried out, including: Binary Classification Labelling: Each image was labelled as Authentic or AI-Generated based on verified source provenance. Source Verification: Authentic images were validated through trusted re

Scope of the Project This project focuses on the design and development of a web-based hybrid AI-image detection system (ProofPulse) capable of distinguishing between authentic and AI-generated images. The system integrates custom-trained deep learning models with external detection services to cross-validate results and present confidence-based classifications through an interactive user interface. The scope covers dataset preparation, model training, system integration, evaluation, and deployment of a user-accessible platform, but does not include real-time social media scraping, video-based deepfake detection, or large-scale cloud production deployment. Specific Data Labelling Tasks To ensure accurate supervised learning, structured data labelling activities were carried out, including: Binary Classification Labelling: Each image was labelled as Authentic or AI-Generated based on verified source provenance. Source Verification: Authentic images were validated through trusted re

2025 - 2025

Education

Work History