Theodoros Vasiloudis

Title: Theodoros Vasiloudis and Graph-Based Entity Representations

Introduction

Theodoros Vasiloudis is an inventor associated with Seattle, Washington, and Amazon Technologies, Inc. His patent work explores how language models can combine information from graphs, text, and images to represent entities.

Latest Patents

Vasiloudis is named on 1 patent: “Systems for Determining Representations of Entities Based on Graph Data Using a Language Model.” The invention describes generating separate representations from graph data, text, and images associated with an entity, then providing those representations to a large language model to produce an overall entity representation. The graph component captures characteristics and relationships between entities. The resulting representation can support other models in identifying entities with similar or differing characteristics.

Career Highlights

The available information identifies Vasiloudis as an inventor working at Amazon Technologies, Inc. His patent reflects an approach to combining multiple data types for entity representation and downstream analysis.

Collaborations

The patent information also names Houyu Zhang and Seongjun Yun as Vasiloudis’s coworkers.

Conclusion

Theodoros Vasiloudis’s patent focuses on using language models to integrate graph, text, and image information into entity representations that can support comparisons and other analytical tasks.