Dhia H.

- Built AI systems for structured information extraction from documents (ID cards, licenses, etc.), processing over 6 million documents with PySpark.

- Designed reusable document processing packages (some open-source) to accelerate the development of document processing engines.

- Automated the indexing of 70% of AXA France's incoming documents using retrainable AI systems for document classification.

- Developed an LLM-based RAG system to answer insurance agents' questions and reduce support team workload.

- Reduced the workload of the support team by developing an LLM-based RAG system to answer insurance agents' questions.

- Built reusable Azure ML pipelines for continuous training (CT), facilitating model deployment in the Model Registry.

- Designed Azure DevOps CI/CD templates, improving code quality and project delivery time.

- Supervised the recruitment process, onboarding new team members, and facilitating effective knowledge transfer.

- Coordinated the annotation process with business stakeholders to ensure high-quality datasets across various projects.

- Led AXA's Python Community of Practice by organizing monthly presentations on topics such as testing, web scraping, and code quality.

- Conducted training sessions on data manipulation with Python for over 30 AXA employees, enhancing their skills

- Built an AI-driven fraud and abuse detection tool, resulting in savings exceeding €150,000 in 2022.

- Designed and trained multiple fraud detection models, significantly improving detection performance.

- Applied NLP techniques to cluster claims adjusters' notes, revealing emerging fraud patterns through advanced visualization.

- Developed and deployed new AI models for Shift's fraud detection solution, enhancing customer offerings.