Ikrame Atir

Key Skills

Software

Label Studio

OpenCV AI Kit (OAK)

Roboflow

Top Subject Matter

No subject matter listed

Top Data Types

Document

Image

Medical Dicom

Text

Top Label Types

Bounding Box

Classification

Data Collection

Entity Ner Classification

Evaluation Rating

Fine Tuning

Object Detection

Relationship

Text Generation

Text Summarization

Freelancer Overview

I am a computer science and networks engineering student with hands-on experience in data labeling, annotation, and AI training data pipelines. I have worked on computer vision and NLP projects where I fine-tuned models such as YOLO and LayoutLMv3, using tools like Label Studio and Roboflow for dataset annotation, cleaning, and augmentation. My experience includes structuring unstructured data, automating data extraction from documents, and evaluating model performance using metrics like ROUGE. I am skilled with Python, Pandas, NumPy, OpenCV, and Hugging Face Transformers, and have applied these in real-world applications such as equipment extraction from brochures, document intelligence for recruitment, and license plate recognition. I am passionate about ensuring data quality to improve AI model outcomes and enjoy tackling challenges in both vision and language domains.

Labeling Experience

AI Apprentice – Computer Vision & PDF Processing

Label StudioDocumentEntity Ner ClassificationClassification

Designed an equipment extraction system from earthmoving excavator brochures, automatically identifying equipment and its status. • Fine-tuned a YOLO vision model on a custom dataset annotated with Label Studio; performed dataset cleaning, structuring, and improvement. • Automated the generation of an Excel report listing all detected equipment from model predictions. • Implemented a RAG module to normalize equipment naming: searches an internal database for equivalences between competitor and standard Liebherr terms, eliminating duplicates.

Designed an equipment extraction system from earthmoving excavator brochures, automatically identifying equipment and its status. • Fine-tuned a YOLO vision model on a custom dataset annotated with Label Studio; performed dataset cleaning, structuring, and improvement. • Automated the generation of an Excel report listing all detected equipment from model predictions. • Implemented a RAG module to normalize equipment naming: searches an internal database for equivalences between competitor and standard Liebherr terms, eliminating duplicates.

2025

AI Engineering Intern – RAG Chatbot & Document Analysis

RoboflowTextEntity Ner ClassificationClassification

Developed a chatbot based on Nemotron (NVIDIA), capable of answering user questions in natural language. • Collected, cleaned, and structured unstructured internal data (PDF, DOCX, emails), then integrated it into the system via a RAG approach combining document retrieval and response generation. • Designed a user interface to facilitate access to the AI assistant.

Developed a chatbot based on Nemotron (NVIDIA), capable of answering user questions in natural language. • Collected, cleaned, and structured unstructured internal data (PDF, DOCX, emails), then integrated it into the system via a RAG approach combining document retrieval and response generation. • Designed a user interface to facilitate access to the AI assistant.

2024 - 2025

AI Engineering Intern – NLP & Document Intelligence

RoboflowDocumentRelationshipFine Tuning

• Contributed to an AI-based recruitment solution. • Developed proofs of concept in AI applied to complex document processing (CVs, cover letters...). • Used Python, LangChain, and TypeScript to develop AI algorithms and manipulate data. • Fine-tuned LLMs (PaliGemma, LayoutLMv3) to automatically extract information from structured documents. • Evaluated models with LangSmith and the ROUGE metric to measure prompt and response quality.

• Contributed to an AI-based recruitment solution. • Developed proofs of concept in AI applied to complex document processing (CVs, cover letters...). • Used Python, LangChain, and TypeScript to develop AI algorithms and manipulate data. • Fine-tuned LLMs (PaliGemma, LayoutLMv3) to automatically extract information from structured documents. • Evaluated models with LangSmith and the ROUGE metric to measure prompt and response quality.

2024 - 2024

Education

Work History