Ayman Knadel

Key Skills

Software

LabelImg

Label Studio

Prodigy

SuperAnnotate

Roboflow

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Image

Text

Top Task Types

Audio Recording

Bounding Box

Classification

Object Detection

Text Generation

Freelancer Overview

Freelance AI engineer with solid experience in data annotation for NLP and computer vision tasks. I’ve contributed to training AI models by labeling visual and textual datasets using tools like LabelImg, Label Studio, and Prodigy. In a computer vision project, I supported the development of plant species detection algorithms by annotating drone footage, optimizing labeling workflows for improved model accuracy. I also led an NLP project aimed at detecting mental health conditions from speech-derived text. I manually labeled large datasets with indicators of depression, anxiety, and suicidal ideation, then trained multi-label classifiers using scikit-learn, TF-IDF features, and Logistic Regression. This work combined precise annotation, domain knowledge, and model evaluation expertise.

Labeling Experience

Plant Species Detection from Aerial Drone Footage

LabelimgVideoBounding BoxObject Detection

Created high-quality bounding box annotations for plant species (invasive vs. native) in drone-shot videos as part of an AI-based environmental monitoring project. Used LabelImg and Label Studio, optimized the labeling pipeline, and contributed to the training of MASK RCNN and YOLOv5 detection models.

Created high-quality bounding box annotations for plant species (invasive vs. native) in drone-shot videos as part of an AI-based environmental monitoring project. Used LabelImg and Label Studio, optimized the labeling pipeline, and contributed to the training of MASK RCNN and YOLOv5 detection models.

2025

Data Annotation & QA Intern – VisionWorks AI

RoboflowImagePoint Key PointEmotion Recognition

Labeled over 80,000 facial images for facial landmark positioning and emotion classification as part of a smart city surveillance dataset. Developed a custom OpenCV-based QA script to automatically detect misaligned keypoints and flag edge cases. Reduced overall annotation errors by 37% through active feedback loops with the ML team. Contributed to protocol documentation and collaborated in maintaining annotation consistency across multiple contributors.

Labeled over 80,000 facial images for facial landmark positioning and emotion classification as part of a smart city surveillance dataset. Developed a custom OpenCV-based QA script to automatically detect misaligned keypoints and flag edge cases. Reduced overall annotation errors by 37% through active feedback loops with the ML team. Contributed to protocol documentation and collaborated in maintaining annotation consistency across multiple contributors.

2024 - 2024

Junior NLP Annotator

SuperannotateAudioText Generation

As part of a multilingual AI training initiative at LangStudio Africa, I contributed to the development of a Moroccan dialect chatbot by annotating a large corpus of real customer support dialogues. The project aimed to enhance intent classification and named entity recognition (NER) for conversational AI models adapted to the North African market.

As part of a multilingual AI training initiative at LangStudio Africa, I contributed to the development of a Moroccan dialect chatbot by annotating a large corpus of real customer support dialogues. The project aimed to enhance intent classification and named entity recognition (NER) for conversational AI models adapted to the North African market.

2023 - 2023

Multilingual Mental Health Text Annotation for NLP Classifiers

ProdigyTextEntity Ner ClassificationClassification

Labeled thousands of speech-derived text samples with psychological indicators such as depression, anxiety, and suicidal ideation. Built annotation guidelines and used Prodigy and custom Python scripts to streamline the process. Leveraged this data to train and evaluate multi-label NLP classifiers using scikit-learn, TF-IDF, and Logistic Regression.

Labeled thousands of speech-derived text samples with psychological indicators such as depression, anxiety, and suicidal ideation. Built annotation guidelines and used Prodigy and custom Python scripts to streamline the process. Leveraged this data to train and evaluate multi-label NLP classifiers using scikit-learn, TF-IDF, and Logistic Regression.

2022 - 2022

Education

Work History