Dr. Chemseddine Berbague

As an R&D engineer at LIPADE, my role encompassed the following responsibilities: • Developing a production-grade pipeline leveraging temporal RDF knowledge graphs to detect and ground factual claims in LLM outputs; covers entity linking, SPARQL temporal triple validation, KG embedding scoring (TransE/RotatE), and precision benchmarking — stack: Python, RDFLib, SPARQL, NetworkX, PyTorch

Co-developed an AI platform for Mediterranean water and environmental management: built a RAG module over geospatial/textual corpora, a multimodal LLM-powered Q&A interface for non-expert users, and a soil impermeability analysis framework — stack: Python, LangChain, FAISS, multimodal LLMs, geospatial APIs

Built and evaluated end-to-end NLP/ML and IR pipelines, benchmarking with NDCG/MAP (Python, spaCy, Hugging Face, NLTK). Led the conversion of the Algerian Official Journal into an OWL/RDF knowledge graph using NER, relation extraction, Protégé, Apache Jena, and SPARQL. Collaborated with KABAS (2024–2025) on agentic AI for virtual labs, covering requirements, automation architecture, model benchmarking, and production readiness.

As an Engineer and Business Analyst, I collaborated with a team to assess the company's needs and propose effective solutions for automating information processes. Through my project work, I gained valuable experience in Python, XML, business analysis, and teamwork. Specifically, my team and I focused on: • Accounting Module Development. • Configurable Reporting Solutions. • Data Verification Automation. • User Interface Improvement.

- Engineered multi-objective recommender systems on large-scale e-commerce data; enriched ranking models with NLP signals (TF-IDF, sentence embeddings) from product metadata and user reviews - Delivered +15% diversity improvement with no accuracy trade-off; designed session-aware pipelines capturing short- and long-term user behavior signal

As a freelance developer, I specialized in creating dynamic web applications and desktop applications using a variety of modern technologies. My experience included: • Frameworks. • State Management. • Database Integration. • Java.

Thesis Title: Studying the Recommendation Novelty and Diversity in Collaborative-based Filtering Approaches. – Summary: The objective of our proposal is to improve the quality of recommendation from different aspects by developing a scalable optimization algorithm that deals with the conflicting combination of different recommendation quality measurements: diversity, novelty, and relevancy.