Raja Sfeir

Key Skills

Software

Data Annotation Tech

Remotasks

Scale AI

Top Subject Matter

No subject matter listed

Top Data Types

Audio

Document

Image

Top Task Types

Classification

Text Generation

Text Summarization

Translation Localization

Freelancer Overview

Most of my work is related to Ai Models Training. I was working with two versions of the same model, and I had to write prompts that belonged to specific categories: Classification prompts, Brainstorming, etc. I worked mostly at DataAnnotationTech, where I submitted around 70 hours of work. I also worked on image annotation, where I had two different projects: In the first one, I had to transcribe the elements in the image, and compare my transcription with the model's. In the second one, I had to find images related to sensitive subjects, and classify the prompt + image by levels of dangerosity.

Labeling Experience

Audio Transcription and Classification

Data Annotation TechAudioClassificationQuestion Answering

In this project, many audio files were submitted where people were asking the model for multiple things, but mostly general knowledge questions and asking for advice. I had to classify the prompts based on: the emotions felt in the voice in the audio, the tone, the speed, and the question itself. Then, I had to give instructions to the AI model, so that it could answer the audio prompt while taking into consideration the emotions identified previously. The goal is here is for the AI model to adapt its responses based on multiple parameters, other than a basic transcription of the question into words.

In this project, many audio files were submitted where people were asking the model for multiple things, but mostly general knowledge questions and asking for advice. I had to classify the prompts based on: the emotions felt in the voice in the audio, the tone, the speed, and the question itself. Then, I had to give instructions to the AI model, so that it could answer the audio prompt while taking into consideration the emotions identified previously. The goal is here is for the AI model to adapt its responses based on multiple parameters, other than a basic transcription of the question into words.

2025

Image annotation

Data Annotation TechImageClassificationQuestion Answering

The AI was given prompts that referred to certain images. Some prompts were malicious while the images were normal, and vice-versa. The goal was to train the AI to answer correctly to these dangerous prompts, with a polite declining sentence for instance.

The AI was given prompts that referred to certain images. Some prompts were malicious while the images were normal, and vice-versa. The goal was to train the AI to answer correctly to these dangerous prompts, with a polite declining sentence for instance.

2025

Education

Work History