Description

Expert en Intelligence Artificielle (Machine Learning, Deep Learning, Traitement automatique de la langue, Data Science).

Prototypes, supervision des développements, conseil et veille technologique, montage de projet de recherche, communication scientifique, crédit impôt-recherche, j'interviens tout au long du cycle de R&D.

Languages

French
Fluent
English
Fluent
Arabic
Native or bilingual

Workplace preferences

Can work on-site

Paris (up to 50km), Massy (up to 30km), Paris (up to 20km)

Huawei
Senior Researcher NLP
E-COMMERCE
September 2020 - Today (5 years and 9 months)
Helsinki, Finland
Senior Researcher NLP - Building Question-Answering System with Llama-2–7b Model and RAG (Retrieval Augmented
Generation)
- Conducted research and experiments on Hate Speech Detection project (offensive, adult, discrimination, and terrorism).
- Automatically generated offensive data using OpenAI API.
- Implemented advanced multilingual solutions for detecting online offensive content by improving
Pre-Trained Multilingual Models with Vocabulary Expansion.
- Implemented multilingual solutions for detecting online adult content. The model is based on
fine-tuning LLaMA 2 with LoRA.
- Participed on share tasks: OSACT 2020, HASOC 2022, and HASOC 2023, SemEval 2022.

LLM chainlit langchain chromadb Llama 2
Aquila Data Enabler
Senior Applied Scientist NLP
E-COMMERCE
March 2020 - September 2020 (7 months)
Paris, France
Participated in the implementation of a question-answering system using a large database of PDF
documents. The implemented model is based on fine-tuned BERT and TAPAS transformers.
BERT TAPAS
Epita
Analyse des sentiments, Chatbot, Reconnaissance d'entités nommées
SOCIAL NETWORKS
October 2017 - Today (8 years and 8 months)
Paris, France
Worked on different projects: detecting emotions in textual conversations, dialect identification and
sentiment analysis.
+ Developed a system for classifying textual dialogues into emotion classes such as Angry, Happy, Sad,
and Others. Utilized various deep neural network techniques including Recurrent Neural Networks
(LSTM, B-LSTM, GRU, B-GRU), Convolutional Neural Network (CNN), and Transfer Learning (TL).
+ Developed a system for classifying comments into one of 26 classes, corresponding to various dialects
of Arabic language. The implemented system based on Recurrent Neural Networks (BLSTM, BGRU)
using hierarchical classification. We started with a higher level of classification (8 classes) and then
performed the finer-grained classification (26 classes).
+ Developed a system for classifying tweets into one of seven classes, corresponding to various levels
of positive and negative sentiment intensity, that best represents the mental state of the tweeter. The
main idea of our solution is to use transfer learning, which allows to avoid learning from scratch.
Indeed, we start to train a first model to predict if a tweet is positive, negative or neutral. For this
we use an external dataset which is larger and similar to the target dataset. Then, the pre-trained
model is re-used as the starting point to train a new model that classifies a tweet into one of the seven
various levels of sentiment intensity.
Keras Deep Learning Python NLTK