Projet 1: Development of the Back-End for an Application Aimed at Evaluating and Improving Text Clarity:- Development of a linguistic engine to process documents in pivot format and analyze text using traditional NLP techniques. - Integration of various large language models (LLMs), such as GPT-4, Gemini, and Claude 3, into the application, with an easy configuration to switch between models and adjust key parameters, like temperature. - Enhancement of lexicons through data generation using GPT-4 with advanced prompting strategies. - Optimization of application performance and implementation of unit tests, addition of logs, and writing of code documentation to ensure quality and maintainability.Projet 2 : Development of an application for Document Import, Data Extraction, and Interactive Querying:- Designed to import documents (HTML, PDF, etc.), extract relevant data, and enable interactive questioning based on the document’s content using Large Language Models (LLM) and Retrieval-Augmented Generation (RAG).- Focused on robust data extraction, parsing, and vectorization of various content types, such as text, tables, and images/graphs, with a specialized approach for each.Projet 3: Automation of LinkedIn Profile Matching for Recruitment Campaigns using generative AI

Development of tools to automate and streamline the research process:- Developed Python tools for automatic transcription of audio corpora into text using the open-source speech recognition model Whisper.- Implemented automatic annotation using forced alignment of text (sentences, words, and phonemes) with audio.- Designed methods and protocols for data or corpus collection.- Analyzed data or corpora and interpreted results in collaboration with the research team.

Improvement of Orange's Speech Recognition System in a Multi-Accent / Multilingual Context:- Extraction and visualization of acoustic features.- Generation of accents using the unsupervised method Wav2vec 2.0 integrated with Hugging Face.- Enrichment of the system's lexicon by integrating accents generated by the Wav2vec 2.0 model.- Training models on GPUs and conducting in-depth analysis of predictions.

Computational Modeling of Large-Scale Data to Study Linguistic Evolution:- Feature engineering of data based on linguistic knowledge.- Prediction of linguistic variations using machine learning algorithms such as Random Forest.- Calculation of the weight of each variable in the prediction.

Acoustic and Statistical Analysis of Speech Production in Noisy Environments:- Led a team responsible for collecting and annotating spoken language data for our study.- Conducted acoustic data analysis using R and Matlab software, extracting relevant measurements and features for our research.- Collaborated with the research team to interpret the results of data analysis and integrate them into our research process.- Authored scientific papers to communicate our findings, results, and conclusions in specialized academic journals, contributing to the advancement of research in linguistics and acoustics.

University Teaching Experience:Master 1 - Comparative Linguistics Université Paris 7 - Denis DiderotMaster 1 - Phonology Université Paris 7 - Denis DiderotBachelor 3 - Phonetics 2Université Paris 7 - Denis DiderotMaster 1 - Experimental PhonologyUniversité Paris 7 - Denis Diderot