Imane Momayiz, Developer in Paris, France
Imane is currently unavailable

Imane Momayiz

Bio

Imane is a machine learning engineer with 4 years of experience, specializing in RAG systems, LLM fine-tuning, and generative AI. She has built production retrieval pipelines and agentic systems for multiple industries, and has worked across the full ML application lifecycle. Imane is fluent in English, French, and Arabic.

Portfolio

Illuin Technology
BentoML, Embedding Models, Java, Python, Milvus, Visual Language Models (VLMs)...
Equancy
Scikit-learn, Polars, Pandas, Streamlit, Forecasting, Clustering...

Experience

  • Python - 5 years
  • Machine Learning - 4 years
  • Deep Learning - 3 years
  • Docker - 3 years
  • Google Cloud - 3 years
  • RAG Pipelines - 2 years
  • Large Language Models (LLMs) - 1 year
  • Agentic AI - 1 year

Preferred Environment

Python, Google Cloud, LangChain, Transformers, Hugging Face, PyTorch, Apache Airflow, Scikit-learn, Docker, Git

The most amazing...

...thing I've built is a production RAG pipeline powering a conversational platform used by millions, boosting retrieval accuracy by 36% for more reliable search.

Work Experience

Senior Machine Learning Engineer

2025 - 2026
Illuin Technology
  • Led the research and redesign of the RAG pipeline powering the company's core conversational product used by millions of users, building synthetic benchmarks and an evaluation framework to compare retrieval strategies systematically.
  • Tested state-of-the-art retrieval and parsing techniques (late-interaction multi-vector, rerankers, VLMs), achieving a 36% gain in retrieval accuracy across real-world and synthetic corpora.
  • Collaborated with product to define the AI feature roadmap, shaping priorities for search and agentic capabilities, and owned deployment to production with the ops team.
  • Contributed to "Compass," the company's external weekly genAI newsletter, sharing applied research with the community.
Technologies: BentoML, Embedding Models, Java, Python, Milvus, Visual Language Models (VLMs), RAG Pipelines, Benchmarking, Large Language Model Operations (LLMOps), ChromaDB, Prompt Engineering

Senior Data Scientist

2022 - 2025
Equancy
  • Built operational forecasting models for a transportation firm covering 100+ targets, used Airflow for batch predictions and DASH for a simulation app.
  • Developed forecasting models to enhance sales predictions and optimize promotion strategies for a major retail company, surpassing their baseline by over 7%.
  • Built a MultiModal RAG to assist technicians during their interventions. Fine-tuned Stable Diffusion models for customized image generation.
  • Designed customer segmentation models for a leading cosmetics company, improving targeting and reducing churn. Developed clustering and lookalike models for a hotel chain, resulting in a 4-point increase in CTR.
  • Lead company-wide effort to implement a multimodal RAG framework.
  • Built scoring models for a major Swiss coffee producer using XGBoost to optimize re-purchase behaviors and promotion sensitivity, delivering actionable insights for managing their B2C portfolio.
Technologies: Scikit-learn, Polars, Pandas, Streamlit, Forecasting, Clustering, Client Communication, Classification, Python, Git, Docker, Data Analysis, Time Series

Experience

AtlasOCR

https://arxiv.org/abs/2604.08070
The first open-source OCR model for Moroccan Darija, an under-resourced language with no standardized writing system.

As the first author, I led the project end-to-end: built OCRSmith, an open-source toolkit for synthesizing tens of thousands of labeled images, and curated real-world data from scanned books, social media, and documents. I fine-tuned a 3B vision-language model (Qwen2.5-VL) using QLoRA, and released AtlasOCRBench, the first evaluation benchmark for Darija OCR. Despite its Darija focus, the model generalizes to standard Arabic, competing with much larger models like Gemma 3 (12 billion parameters) and Qwen2.5-VL (7 billion parameters) on the KITAB-Bench benchmark.

Earthquake Assistance Collaborative Platform

https://huggingface.co/nt3awnou
Took part and led the data team in building a platform to coordinate relief efforts for individuals impacted by the 2023 Morocco earthquake. The platform received over 2,000 requests, facilitating more than 300 interventions.
The platform ranked 5th among the most viewed spaces on Hugging Face during the peak period, and the paper was accepted at NeurIPS NAML 2023.

Education

2018 - 2021

Master's Degree in Mathematics and Computer Science

IMT Atlantique - Brest, France

Certifications

DECEMBER 2023 - DECEMBER 2025

Google Cloud Professional Machine Learning Engineer

Google Cloud

Skills

Libraries/APIs

Pandas, Scikit-learn, PyTorch, BentoML

Tools

Git, Visual Language Models (VLMs), Claude, Apache Airflow

Languages

Python, Java

Platforms

Docker, Vertex AI, Kubernetes

Storage

Google Cloud

Frameworks

Streamlit

Paradigms

Synthetic Data Generation

Other

LangChain, Transformers, Hugging Face, Machine Learning, Deep Learning, Embedding Models, RAG Pipelines, Agentic AI, Large Language Models (LLMs), Milvus, Benchmarking, Data Collection, LoRa, Qwen, Fine-tuning, Supervised Fine-tuning (SFT), Training, Open-source LLMs, Google BigQuery, Blob Storage, Polars, Forecasting, Clustering, Client Communication, Classification, Data Analysis, Time Series, Large Language Model Operations (LLMOps), ChromaDB, Prompt Engineering

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring