Raven is an AI based system developed at JAMK University of Applied Sciences (JAMK) for clustering and visual examination of scientific abstracts in large datasets. Instead of relying on traditional keyword based search or manual classification, Raven infers the semantic proximity between abstracts directly from the text, which enables the examination of tens or even hundreds of thousands of documents as an interactive, content structured landscape. Because the models are run locally, sensitive or non public material can be analyzed without the data leaving the organization’s infrastructure; this supports privacy preserving literature reviews and internal foresight/horizon scanning. We demonstrate Raven with a case study that used abstracts from the ECCOMAS conference held in Lisbon and examined how contributions are distributed across different themes. The resulting maps enable users (i) to grasp the breadth and balance of a research area at a glance, (ii) to discover under explored sub areas adjacent to their own interests, and (iii) to rapidly assemble representative document sets for deeper analysis. Raven is open source software: https://github.com/Technologicat/raven