This inventor holds 1 USPTO granted patent and 1 published patent application. Top assignee: Oracle International Corporation. Active years: 2026.
Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile
Company Filing History:
Years Active: 2026
Title: Marcel Gwerder and Synthetic Tabular Metadata Generation
Introduction
Marcel Gwerder is an inventor based in Dietikon, Switzerland. His patent work includes a method for generating synthetic tabular metadata with large language models. He is associated with Oracle International Corporation and has 1 patent listed in the information provided.
Latest Patents
Gwerder’s patent, “Synthetic Tabular Metadata Generator Using Large Language Models,” describes a computer-generated lexical prompt that guides a large language model in inferring a hybrid table schema containing natural-language descriptions of a data table. The prompt can include linguistic examples drawn from predefined tables, selected either statically or dynamically.
The described approach uses static examples to improve task accuracy and dynamically selected, semantically similar examples to improve semantic accuracy. It also uses indexed semantic vector encodings of existing and new data tables to accelerate the selection of dynamic examples.
Career Highlights
The available information identifies Gwerder’s work with Oracle International Corporation and highlights his contribution to applying large language models to tabular metadata generation. The patent describes techniques intended to improve the accuracy and efficiency of computer inference.
Collaborations
Clemence Lanfranchi and Margherita Musumeci are identified as Gwerder’s coworkers. The information provided does not specify their roles in the patent or the invention.
Conclusion
Marcel Gwerder’s listed patent presents an approach to generating table schemas with large language models, combining carefully selected examples and semantic indexing to support more accurate and efficient inference.
Data source: USPTO (United States Patent and Trademark Office) public patent records. Weekly synchronization. How IDiyas builds this profile
