Amjad Ben Essafi

Key Skills

Software

Labelbox

Top Subject Matter

No subject matter listed

Top Data Types

Computer Code Programming

Text

Top Task Types

Computer Programming Coding

Text Generation

Freelancer Overview

With extensive experience in AI training and data labeling, I specialize in evaluating and improving AI-generated code, as well as developing and refining testing pipelines for robust model performance. I have worked on creating adversarial prompts to challenge and enhance the robustness of models like Nova and GPT-4, achieving a 100% success rate in CI/CD validation. My expertise spans across multilingual programming in languages like Python, C/C++, PHP, and JavaScript, enabling me to provide precise and efficient solutions for AI system development and model evaluation.

Labeling Experience

Adversarial Prompt Engineering & CI/CD Validation for AI Model Evaluation

LabelboxComputer Code ProgrammingEvaluation RatingRed Teaming

In the CODE VERIFIER Project, I engineered over 50 adversarial prompts that consistently challenged and failed models like Nova and GPT-4. These prompts were carefully designed to test the limits of the models and uncover weaknesses in their output, ensuring robust performance under edge cases. Additionally, I implemented an automated GitHub CI/CD pipeline using Python, Pytest, and GitHub Actions to verify pull requests (PRs), achieving a 100% success rate in PR validation and over 90% test coverage for reliable code verification. The project involved working closely with Large Language Models (LLMs), generating a wide range of test scenarios and ensuring the code verification process was seamless and efficient. The focus was on creating data that could evaluate AI performance and produce insightful results for training improvements.

In the CODE VERIFIER Project, I engineered over 50 adversarial prompts that consistently challenged and failed models like Nova and GPT-4. These prompts were carefully designed to test the limits of the models and uncover weaknesses in their output, ensuring robust performance under edge cases. Additionally, I implemented an automated GitHub CI/CD pipeline using Python, Pytest, and GitHub Actions to verify pull requests (PRs), achieving a 100% success rate in PR validation and over 90% test coverage for reliable code verification. The project involved working closely with Large Language Models (LLMs), generating a wide range of test scenarios and ensuring the code verification process was seamless and efficient. The focus was on creating data that could evaluate AI performance and produce insightful results for training improvements.

2025 - 2025

Education

Work History