Pierre de La Ruffie

🍎 My work as a member of the SIML branch revolved around quantization, post-training optimization and ML model size reduction while preserving accuracy. I also worked on the integration of an external optimization pipeline into Apple's internal infrastructure, and support on an internal platform for optimized, distributed ML model training.

🏢 Company Presentation: Datakalab is a startup founded in 2016. Initially focused on R&D in artificial intelligence, the company later specialized in AI model compression and their deployment on various hardware for on-edge inference. 💼 Quantization and Model Reduction: My work involved implementing new quantization methods in an existing application, maintaining it, and using it to reduce the size of client models. Additionally, as part of the Confiance AI initiative, I contributed to a joint project with IRT Saint-Exupéry.

🏢 Company Presentation: Dataswati is a startup founded in 2016, specializing in optimizing industrial processes through artificial intelligence, data analysis, and machine learning. Its only product to date, PowerOp, is a web application designed to manage, analyze, and aggregate industrial production data. 💼 Internship: The first part of the internship focused on acquiring new concepts that had not been covered in my academic curriculum: Front-end development with the Vue.js framework, communication with a Symfony back-end in PHP, anonymized databases, and project management within a professional team. The remainder of the internship involved assisting the development team in maintaining the Front-end of the PowerOp tool, as well as developing new features.

The internship primarily involved using image deployment strategies on a computer network with several hundred units, utilizing FOG. This system also included leveraging specific features of the Windows operating system (notably the registry) to configure multiple machines repeatedly using scripts.