Data Scientist / AI Engineer
Ayoub El Majjodi
I build search, recommendation, and personalization systems, currently focused on ranking and discovery in the real estate domain, and on LLM- and agent-based systems that make search and recommendations more helpful.
Previously a PhD researcher in recommender systems and behavioral data science at the University of Bergen, studying how personalization shapes decisions and encourages healthier behavior.
About
I'm a Data Scientist and AI Engineer with a PhD in Information Science from MediaFutures and the University of Bergen. I'm passionate about AI systems that bridge research and real-world applications. My interests span LLMs, search, personalization, retrieval, and agentic AI, with a focus on evaluation, benchmarking, and scalable AI engineering.
I enjoy working at the intersection of AI, engineering, and product, collaborating across disciplines to build reliable AI-powered experiences. Today that means LLM-based property understanding, agentic search and discovery workflows, and rigorous benchmarking of embedding models, retrieval strategies, and LLMs in the real estate domain.
My PhD was supervised by Prof. Christoph Trattner and Assoc. Prof. Alain D. Starke, within the Behavioral Data Analytics & Recommender Systems group and Work Package 2, User Modeling, Personalization & Engagement, at MediaFutures. It explored how personalization and recommender systems influence decision-making and promote healthier behavior across the food and news domains.
Experience
Build and evaluate AI capabilities for search, discovery, and personalization that improve relevance and engagement in the real estate domain. Design agentic search and retrieval workflows (LangGraph, LangSmith); run systematic benchmarking of embedding models, retrieval strategies, LLMs, and document-parsing pipelines; and partner with product and design to define, prototype, and validate AI features across several initiatives.
Smart Search, a natural-language search experience for dinbostad.se. Users describe what they are looking for in plain text; the system infers intent with LLMs and returns relevant listings. Scope spanned LLM selection, prompt engineering, retrieval design, and end-to-end evaluation and benchmarking of the pipeline.
Recommender systems, user modeling, and behavioral interventions across the food and news domains; offline and online evaluation of personalization.
Education
Expertise
Credentials
Service
Reviewing full and short papers on recommender systems, personalization, and user modeling.
Coordinated the camera-ready proceedings and the conference video program with authors and publishers.
Supported organisation, sessions, and participants at the ACM Recommender Systems Summer School.