Dhia Abdelli

• Design and implementation of a Graph-based machine learning interface designed to help R&D engineers build and iterate on AI workflows with minimal coding • Integration of large language models with tabular foundation models to enable insight extraction from structured engineering data • Implementation of modular components for data preprocessing, model training, and evaluation to support flexible ML workflows

• Development of an ML pipeline to estimate hydrogen concentration from existing sensor data, replacing dedicated hardware • Integratation of MLOps components to ensure reproducibility and model traceability • Deployment of the first embedded AI model for real-time execution in fuel cell systems

• Contributed to the identification and development of business-relevant data science and machine learning use cases, collaborating with cross-functional teams to ensure alignment with organizational goals. •Built and implemented end-to-end machine learning projects, leveraging inline process parameters and measurement equipment, providing data-driven insights for process and product analysis.

•Conducted in-depth experiments of different statistical and neural network-based models for keyword extraction and benchmarked them on public datasets •Optimized relevance and diversity of results by combinig TF-IDF and KeyBERT. •Engineered a microservices architecture for "Kiwrd," containerizing the different micro services which significantly improved scalability, resulted in low-latency (20% faster compared to a monolithic architecture) and enhanced fault tolerance.

•Discovered meaningful correlations between various store performance indicators, such as sales volume and customer satisfaction. •Utilized statistical techniques to elucidate how specific factors, such as store location or product placement, influenced the overall performance of each shop •Developed a comprehensive sales insights dashboard using PowerBI and Django for a web alternative, providing shop owners with an accessible and intuitive tool for data-driven decision-making.