Constantin, AuroraAlsalman, Safa2023-11-132023-11-132023-11-13https://hdl.handle.net/20.500.14154/69640This project investigates developing a tool to generate academic profiles from multiple sources to tailor to the needs of Business Development Executives (BDEs) and their external partner. a User-Cenered Design (UCD) approach was utilized during the lifecycle of the project, emphasizing users’ needs, preferences, and challenges throughout the design process. The methodology was divided into three stages: the pre-design, formative evaluation, and summative evaluation stages. The studies aimed to identify the challenges the BDEs encounter with the current tools, the requirements for the tool and the best way to personalize the experience of building profiles to users. The project also investigated incorporating a Large Language Model (LLM) to generate the biographies automatically and aimed to assess the quality of the generated text with BDEs as human evaluators to judge its clarity, concision, and suitability for audiences with different knowledge base. The results show that the main challenges with the existing tools include the lack of sufficient information on one single profile, the presence of outdated information, the inconsistency in the structure and presentation of the profiles and the lack of a mechanism for maintaining and disseminating the information across BDEs. As revealed by the semi-structured interviews, opinions of BDEs about what information to include in the profile vary. Accordingly, the tool has to allow users to curate the profiles according to their perceptions of what their clients may need. Running informative evaluation to inform the design of the tool’s User Interface (UI) showed that users also differ on the way they prefer for building the profile, with some users preferring to see the whole picture first while others prefer to build the profile step by step. By running a summative evaluation study to assess the usefulness of the tool and its effectiveness in generating accurate and relevant profiles, the findings shows an acceptable level of accuracy. The relevancy aspect was assessed in Think-Aloud sessions with BDEs and external partners which revealed that BDEs thought the standard information was relevant while the external partners were less interested in the overview and more interested in the relevant work. As for the quality of the text of the generated biography by the language model, BDEs expressed a level of satisfaction with one form over the others. The findings of the summative evaluation suggest the tool is perceived as useful by BDEs with an excellent SUS score and overall satisfaction with the features provided.100en-USUser-Centered DesignAcademic ProfilesAutomatic GenerationPre-trained LLMThesis