Khouloude Fathallah

Key Skills

Software

AWS SageMaker

Appen

CVAT

Dataloop

Label Studio

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Image

Text

Top Task Types

Action Recognition

Audio Recording

Evaluation Rating

RLHF

Text Generation

Freelancer Overview

I directly worked on AI training projects where I was required to evaluate audio and text output from AI systems. For audio recordings, I needed to review different aspects including harmfulness, truthfulness, and listening ease to verify whether the content was of high quality. Additionally, I evaluated AI-text against such criteria as completeness (with respect to the given prompt) and groundedeness, ensuring responses were accurate and well-formed. My activity also involved verifying tasks passed to me by other contributors, with a view to offering guidance for quality outputs. This two-stage process of evaluation and reviewing AI-produced work has cultivated my attention to detail and working knowledge of quality control in training AI, qualifying me to offer insightful contributions towards the development of robust AI systems.

Labeling Experience

AI trainer

Scale AIImageClassificationPrompt Response Writing SFT

Evaluating LLMs responses based on images' inputs.

Evaluating LLMs responses based on images' inputs.

2023 - 2024

Ai Trainer

Scale AIAudioEvaluation RatingData Collection

Recording and transcribing a set of audios.

Recording and transcribing a set of audios.

2023 - 2024

Ai trainer

Scale AITextText GenerationText Summarization

The project aimed to improve AI models' ability to generate accurate, factual, and well-supported responses in text, audio, and image formats. For text, I made sure responses are complete, true, and coherent with the prompt. In audio, I checked if responses were human-like, complete, and contextually appropriate. I also checked the models' ability to comprehend and respond to images. In over 11 months I completed over 6,000 tasks and assisted in raising the overall quality of the models. My responsibilities in text data labeling included verifying the validity of claims and making responses comply with prompt demands. In audio, my responsibility was to provide human-like, complete, and relevant responses. In RLHF tasks, I wrote French prompts and verified responses from two AI models against each other and located errors. I worked within strict guidelines and training modules to guarantee top-class standards for the entire project.

The project aimed to improve AI models' ability to generate accurate, factual, and well-supported responses in text, audio, and image formats. For text, I made sure responses are complete, true, and coherent with the prompt. In audio, I checked if responses were human-like, complete, and contextually appropriate. I also checked the models' ability to comprehend and respond to images. In over 11 months I completed over 6,000 tasks and assisted in raising the overall quality of the models. My responsibilities in text data labeling included verifying the validity of claims and making responses comply with prompt demands. In audio, my responsibility was to provide human-like, complete, and relevant responses. In RLHF tasks, I wrote French prompts and verified responses from two AI models against each other and located errors. I worked within strict guidelines and training modules to guarantee top-class standards for the entire project.

2023 - 2024

Education

Work History