Automating Formal Review of Grant Applications

Supervisor: Prof. Dr. Rico Sennrich

Even before grant applications are reviewed on their quality, they need to be checked for their formal correctness: are all required documents provided? Are applicants eligible for the grant, for example in terms of academic age or their current position? 

The goal of this project is to investigate if the formal review of grant applications can be supported by Large Language Models. You will investigate different strategies and model types to handle the multimodal, multilingual nature of application documents (including passports, transcripts etc.). Due to the confidential nature of applications, the project will focus on local open-weight models.

The project will be executed in collaboration with theUZH Research and Grants Office, who will provide data from the UZH Candoc and Postdoc Calls, and outline design requirements. Students are expected to treat data confidentially, and to grant the Research and Grants Office a non-exclusive license to the project results.

• Background in machine learning, especially LLMs
• Good programming abilities (Python)