Maroua Aaboud

– Designed and implemented an enterprise Knowledge Graph using RDF/OWL and SPARQL for data modeling – Built natural language (NL) to SPARQL dataset generation pipeline with automatic query synthesis and paraphrase augmentation – Fine-tuned an LLM with QLORA for NL to SPARQL translation; added validation and fallback strategies to reduce invalid queries – Implemented semantic retrieval with vector embeddings and OpenSearch to support schema grounding and RAG-style answering

– Developed scalable data pipelines for IoT sensor data (vision, Bluetooth) with Dockerized workflows – Deployed AI pipelines on Azure – Designed and deployed a Retrieval-Augmented Generation (RAG) chatbot using Mistral-7B integrated into the company SaaS platform – Contributed to 3D reconstruction module and occupancy heatmap visualization for space analytics

– Researched context in AI and viewpoint logics; proposed a robust extension of OWL 2 with modal operators – Improved ontology expressiveness while maintaining reliable reasoning compared to metadata-based approaches

– Built supervised classification model (logistic regression) to predict IBD-related associations; tuned regularization to reduce overfitting – Implemented an object detection fine-tuning project (ResNet18 / VOC) using PyTorch and Torchvision