Clemence Lanfranchi

Title: Clemence Lanfranchi and Innovations in Large Language Models

Introduction

Clemence Lanfranchi is an inventor based in Zurich, Switzerland, whose patent work explores the use of large language models (LLMs) to improve database-related tasks and generate synthetic data. Lanfranchi is associated with Oracle International Corporation and holds 2 patents.

Latest Patents

One of Lanfranchi’s latest patents, “SQL Fixit—Automated Generation of Fine-Tuning Data Using LLMs,” describes a method for creating fine-tuning data that can improve an LLM’s accuracy when generating database statements. The approach uses a natural-language request to produce an incorrect database statement and a correct alternative, then generates a restatement of the request based on a second database schema. These elements are used as inputs for fine-tuning the target model.

Another patent, “Synthetic Tabular Metadata Generator Using Large Language Models,” describes generating hybrid table schemas that combine structured table information with natural-language descriptions. The method uses linguistic examples drawn from predefined data tables. Static examples can support task accuracy, while dynamically selected, semantically similar examples can improve semantic accuracy. Indexing semantic vector encodings helps accelerate the selection process.

Career Highlights

Lanfranchi’s patent portfolio focuses on applying LLMs to database statement generation, model fine-tuning, and synthetic tabular metadata. Together, the inventions address ways to create useful training examples and improve the accuracy of computer inference.

Collaborations

Lanfranchi is associated with Oracle International Corporation. The provided information also names Marcel Gwerder and Margherita Musumeci as coworkers.

Conclusion

Clemence Lanfranchi’s two patents highlight practical approaches to using large language models for database tasks, fine-tuning data generation, and synthetic table metadata.