BT/MT A Hybrid User Interface with Agentic AI for Travel Destination Recommendation

The domain of travel and tourism is complex for recommender systems, because the items to be recommended, e.g. travel destinations, have many facets. In previous work, we have developed DestiRec, a Web-based solution for destination recommendation [1] (www.destirec.com). The application allows for interactive specification of preferences (e.g. travel month, activities) and adapts recommended travel regions instantly on a map.

The goal of this Bachelor’s or Master's Thesis is to extend this application with user interaction based on Large Language Models (LLMs) and Agentic AI. The idea is to allow users to specify their travel preferences and constraints in a natural language dialogue and replace the recommendation algorithm with an LLM-based approach that is still grounded in the existing dataset and shows the suitability of travel regions in the map of the DestiRec application. An important feature is the integration of LLMs with map-based visualization and interaction, so the main scientific aspect is the investigation of hybrid user interfaces with agentic AI. Possible extensions include the design of a profile agent that models and manages learned user preferences based on previous trips and other information. Another idea is to allow users to upload images of desired locations and experiences, and investigate how to analyse these examples and integrate them into the recommendation system. The solution should be tested in a user-centered evaluation using appropriate methodology.

It is possible to adapt the topic or split the features among several students and work on the topic in a small team. The prerequisites for this project are very good programming skills and ideally experience in developing agentic and LLM-based applications with the relevant frameworks such as LangChain.

Please send your application (brief CV, transcript of records and short motivation statement including ReactJS experience) to Wolfgang Wörndl (woerndl(at)cit.tum.de) until September 28th, 2025 at the latest, the actual start of the thesis project is flexible.

[1] Asal Nesar Noubari, Wolfgang Wörndl: Dynamic Adaptation of User Preferences and Results in a Destination Recommender System. WSDM 2023 Workshop on Interactive Recommender System (IRS), Singapore, Mar. 2023 [Paper]