- Unsupervised text classification using GPT embeddings and similarity score- Network theory for skill visualisation and content-based recommendation- Metadata enhancement using LLM implementation

Live Media about entrepreneurship and startup ecosystem, from all backgrounds

- Content-based recommendation model using Universal Sentence Encoder & Cosine similarity score - Customer feedback sentiment analysis with supervised polarity model- Generative AI for content enrichment and metadata generation- Data analysis stack maintenance

Data Driven and AI tools development (Public sector customer - Electricity distribution):• Dashboard tool for customer feedback analysis (Approx. 10k+ weekly comments for inference):‣ Machine Learning: State-of-the-art NLP classification models development (topic classification + sentiment analysis) using sentence embeddings approaches (Word2Vec, CamemBERT, Transformers architecture)‣ Backend: Python Flask API development for ML deployment, optimized and automated data extraction/preprocessing/insertion pipelines‣ Database: Database architecture definition and setup (PostgreSQL databases)‣ Frontend: Basic frontend development for user interface using R, shiny/shinydashboard frameworks and components‣ Deployment: Container orchestration and deployment with Docker/docker-compose on internal cloudhosting solutions, micro-services distribution with orchestration• Call centers forecasting engine using weather and climate data/forecasts for crisis/rare events (Approx: 150k weekly calls to infer)‣ Machine Learning: Developed call center volume forecasting engine, based on weather forecasts features, using boosting techniques and LSTM approach (R, caret, Xgboost)‣ Frontend: Basic frontend development for user interface using R, shiny/shinydashboard frameworks and components‣ Deployment: Development/production environment setup with DockerData analytics and data science studies for customer knowledge and marketing purposes:• Statistical and semantic analysis of a conversational agent's conversations (Semantic clustering of Word2vec embeddings)• Statistical studies on customer electrical data for marketing analysis and customer relationship managementDashboarding and BI solutions for KPI and metrics visualization for customer electrical-related activities using Tableau software

Call centers forecasting engine using weather and climate data/forecasts for crisis/rare events (Approx: 150k weekly calls to infer):‣ Machine Learning: Developed call center volume forecasting engine, based on weather forecasts features, using boosting techniques and LSTM approach (R, caret, Xgboost)‣ Frontend: Basic frontend development for user interface using R, shiny/shinydashboard frameworks and components‣ Deployment: Development/production environment setup with DockerData analytics and data science studies for customer knowledge and marketing purposes

1st prize in the "research" category of the Cassiopeia programWithin the scope of a school project called, our pair of students accompanied doctoral students' research on machine learning applied to speaker recognition/speech physics. Study focused on speech/voice recognition issues using vocal feature extraction: - Writing of a state of the art and development of an experimental protocol for speaker recognition in medical sector - NN implementation and GMM algorithms for classification using Python Keras frameworks - Experimentation of several configurations to optimize system - Writing and validation of an experimental reportTools used: Python (machine learning library Keras and Scikit Learn)

Data Quality analysis and completion analysis for administered funds data for portfolios managers:• Analysis and understanding of the data architecture and financial information system• Automated data extraction/preprocessing/ingestion pipelines• Basic statistical analysis for data completion and reliability.