The allocation of scientific funding through grant programs is crucial for research advancement. While independent peer panels typically handle evaluations, their decisions can lean on personal preferences that go beyond the stated criteria, leading to inconsistencies and potential biases. Given these concerns, our study employs a novel method, using simulated, data-driven, and narrative personas of fictional candidates, to identify the attributes that significantly influence panelist choices and profile what an optimal candidate would look like. Our findings reveal a preference for mid-career, multidisciplinary researchers with significant publications, citations, and prior project experience. Such optimal candidates also lean toward applied science, collaborative research, interactions with both industry and the public, and a progressive stance on science. Contrary to existing literature, this study found indications of a bias toward female candidates, which we interpret as a result of deliberate correction caused by awareness of existing biases in academia. Age emerged as another influential factor, suggesting either a preference for younger researchers or a perception of waning productivity among seasoned academics. Consistency in a candidate's profile, rather than standout attributes, was also favored by panelists.