Hello! ποΈ I am a Senior AI & Data Science Engineer with 5 years of experience. I excel in designing and implementing data-driven models tailored to client needs.
π Areas of Passion:
- Advanced NLP techniques including:
- LLMs π€
- Qlora π¦
- LORA π¦
- RAG (retrieval-augmented generation) πΈοΈ
- π€Ώ Currently, Iβm immersed in exploring augmented data generation techniques for NLP tasks.
- π¬ Got queries about NLP, AI, or machine learning? Don't hesitate to ask! π§
- Large Language Models (LLMs) π€
- Data Analysis π
- Natural Language Processing π
- Dashboard Realization π
- Business Intelligence π
- Data Management & Transformation βοΈ
- Machine Learning π¦Ύ
- Deep Learning π§¬
- Data Visualization πΌ
Whether you're looking to tell a compelling story with your data, develop a real-time dashboard with KPIs for monitoring your company's health, or explore natural language processing solutions π£, I can assist you in your next venture.
Over the years, I've successfully managed projects worth +β¬1M across 12+ countries π for major clients including the European Parliament, Kering, Atos, Renault Nissan Mitsubishi, Damart, and more. I wear multiple hats as a π§βπ¬ Data Scientist, π Data Analyst, and π§βπΌ Project Manager with both functional and managerial expertise.
- Microsoft Certified: Power BI Data Analyst Associate (PL-300)
- Microsoft Certified: Data Analyst Associate (DA-100)
- Microsoft Certified: Azure Fundamental (AZ-900)
Location: Paris (October 2020)
Prize: β¬10,000
- Objective: Development of a prediction/recommendation platform based on AI.
- Description:
- Prediction of the environmental impact of Kering's various activities throughout the supply chain.
- Accurate evaluation of environmental impacts: resource depletion, biodiversity, greenhouse gases.
- Decision support for designers, material researchers, and consumers in their choices to reduce the impact of luxury.
- Automatic creation of predictive models from user-provided data or directly from Kering data and integrated native models.
- Recommendation technique: Collaborative-based method, Content-based method, Hybrid method.
- Results: The team won 1st place for the best platform for predicting the environmental footprint of Kering's products.
- References:
Location: Paris (February 2019)
- Objective: Production of a Business Intelligence tool based on voice commands.
- Description:
- Real-time monitoring of the actions of each sales agent on the different business units and real-time feedback.
- Automation of the customer prospecting process by centralizing data.
- Streamlining the Manager - Salesperson - Client relationship through an omnichannel approach based on high-value-added artificial intelligence.
- Results: This project allowed decision-makers to discover insights into their activity and monitor the actions of each salesperson in real-time.
- References:
π― Objective: Automate order & credit return management for Thuasne with advanced NLP & LLMs.
π§ Created an email classification system using the Melusine tool.
π Extracted order details like items, quantity using NLP & LLMs.
π οΈ Developed a prototype system handling email text & attachments.
π Improved order processing efficiency & anomaly detection.
π Used explainability tools like LIMETextExplainer, ELI5NLP, SHAP, and AnchorsNLP.
π‘ Optimized system performance with hyperparameter tuning.
π Adapted the system to cater to different orthopedic domains.
π Result: Enhanced system efficiency & performance using advanced NLP techniques, LLMs, and Melusine tool.
π Documentary AI Fraud Demonstrator - Senior Data Scientist NLP
π― Objective: Develop a demonstrator for detecting anomalies, frauds, or inconsistencies in documents.
π Designed a demonstrator for document processing (insurance invoices).
π« Detected fraudulent patterns & document falsification.
π Extracted key invoice fields & verified their consistency.
π΅οΈββοΈ Identified potentially suspicious overbilling cases.
π Result: Significant improvement in fraud detection using AI, outperforming traditional OCR techniques.
ποΈ Part Forecasting: PFO β Technical Lead/ Technical Expert
π― Objective: Provide 18-month forecasts to suppliers, mitigating semiconductor crisis impact.
π» Developed PFO architecture as a Streamlit web app hosted in Azure.
π Integrated data from Oracle Exadata Database.
βοΈ Used Azure Data Factory for file transfer & processing tasks.
π’ Containerized the PFO app using Docker & deployed to Azure Container Registry.
π Result: Enhanced inventory management & supplier collaboration, improving part prediction accuracy.
- Project Management π
- Preparation π
- Planning ποΈ
- Management π
- Evaluation π
- Monitoring and Control of:
- Resources π°
- Calendar π
- Costs πΈ
- Scope π―
- Risk π¨
- Quality π
- Requirements π
- Value π
- Satisfaction π
- Tools: TFS, MS Project, GANT, PERT