Hanen Ben Salem

- Supervised 2 final-year interns (PFE) on an 4-month MSS project management initiative, providing technical mentorship and ensuring on-time project delivery. - Migrated 6 critical Excel files to secure web applications with role-based authentication and access control, eliminating 100% of uncontrolled data-sharing risks. - Automated repetitive workflows with custom-built solutions, reducing manual processing time by 70% and saving teams an average of 15hours/week per employee. - Integrated AI into business processes for automated PDF document generation from forms, cutting document creation time by 80% and improving data accuracy. - Centralized and synchronized data across 3 departments via SharePoint, ensuring real-time data consistency and reducing data duplication by 65%. - Built 2 business applications on Microsoft Power Apps integrated with existing tools, streamlining operations and cutting internal request processing time by 50%. - Developed modern web applications integrated with Microsoft 365 tools, improving cross-team information sharing and reducing internal email exchanges by 40%.

Hybrid Generator of Advanced AI and Quantum for Synthetic Medical Data: Application to Diabetes - Design and development of 3 synthetic medical data generation approaches: rule-based algorithms, classical GAN networks, and quantum GANs (QGAN). - Comparative evaluation of the 3 synthetic datasets against the real MIMIC-III database using advanced statistical metrics and GDPR and HIPAA compliance audit. - Design and deployment of an end-to-end MLOps pipeline with MLflow tracking and real-time monitoring. - Implementation of CI/CD GitHub Actions with DataOps workflow orchestration, continuous deployment, versioned AI model management, and Flask user interface for dataset catalog.

Facial emotion recognition during a job interview: - Balanced RAF-DB dataset across 7 classes, improving model performance and achieving a balanced training set. - Created a CNN model from scratch achieving 92.53% training accuracy and 80.03% test accuracy by combining SMOTE and data augmentation. - Implemented face detection functionality in a Django application, enabling real-time emotion analysis on uploaded images and videos. - Develop a Docker image that includes the application and its emotion analysis feature for easy deployment and scalability.

helped me get this job

Performance of AI models for hate speech detection in Arabic: - Collected and cleaned two sets of data, one with 6,024 Arabic text samples, demonstrating skill in preparing data. - Analyzed six hate speech detection approaches using Weka and Google Colab, comparing models and achieving accuracies between 72% and 83%. - Improve the best model by adding more methods and standard embeddings to make it work even better.

Development of a mobile meal delivery application : - Designed attractive user interfaces with Flutter, guaranteeing a 100% responsive experience. - Improved delivery provider selection for a Saudi client’s app, utilizing Git and increasing efficiency by 20%.