Carlo Musongela

Key Skills

Software

AWS SageMaker

Clickworker

CVAT

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Image

Text

Top Task Types

Question Answering

Text Generation

Translation Localization

Freelancer Overview

I am a versatile Data Labeling and AI Training Specialist with experience working on diverse annotation projects, including text, image, audio, and code-based tasks. Skilled in classification, bounding box annotation, entity recognition (NER), translation, and evaluation, I bring a strong eye for detail and consistency to every dataset. My background in software development (Python, Laravel, Flutter) allows me to understand technical workflows and contribute effectively to projects that require both annotation and technical validation. I have contributed to AI model refinement across multiple domains such as natural language processing (summarization, translation, question answering), computer vision (object detection, segmentation), and LLM evaluation (prompt–response writing, red teaming, and fine-tuning support). Fluent in French, English, and Portuguese, I can work with multilingual datasets and ensure high-quality results that meet project standards.

Labeling Experience

Autonomous Driving Image Annotation

CVATImageBounding BoxSegmentation

Contributed to the annotation of large-scale image and video datasets for autonomous driving systems. Tasks included drawing bounding boxes, segmenting road objects (vehicles, pedestrians, traffic signs), and labeling weather and lighting conditions. Ensured annotation accuracy through double-check quality control (QC) and peer review. Delivered over 25,000+ labeled frames within strict deadlines while maintaining >98% accuracy.

Contributed to the annotation of large-scale image and video datasets for autonomous driving systems. Tasks included drawing bounding boxes, segmenting road objects (vehicles, pedestrians, traffic signs), and labeling weather and lighting conditions. Ensured annotation accuracy through double-check quality control (QC) and peer review. Delivered over 25,000+ labeled frames within strict deadlines while maintaining >98% accuracy.

2024 - 2024

Education

Work History