Yuchen Xia

AI Engineer at Stellantis

Led R&D and deployment of an internal employee chatbot, training the Mistral-7B LLM using PyTorch and Hugging Face Transformers to fully eliminate dependence on commercial APIs and reduce training costs by 100%. Created a domain-specific dataset with over 200 annotated examples updated monthly, enhanced data preprocessing and evaluation pipelines raising accuracy from 87.5% to 100%, validated by human assessors. Collaborated with IT and cloud teams to publish and partially deploy a beta version of the chatbot to internal staff and provided presentations to explain the chatbot functionality to non-technical users.

System Engineer at Zephalto

Researched and prototyped semi-automatic electronic systems for monitoring, adjusting, and optimizing oxygen concentration using real-time sensors. Designed electronic circuits, printed circuit boards, and system architecture with KiCAD. Conducted thorough system requirements analysis, identified technical constraints, and prepared detailed and specific specifications. Manufactured, integrated, and tested prototypes including performance validation and the implementation of optimization processes.

Research Assistant at LAAS-CNRS

Replicated the federated learning system as described in a research paper. Examined the impact of algorithms on machine learning accuracy, provided optimization solutions, and compared learning methods in federated architectures. Optimized classification and algorithm accuracy.

AI Engineer at Palais eauAI engineer | Stellantis

Worked on R&D and deployment of an internal employee chatbot. Trained the Mistral-7B large language model using PyTorch and Hugging Face Transformers, eliminating dependence on commercial APIs and reducing training costs by 100%. Created a domain-specific dataset with over 200 annotated examples, updated monthly based on employee needs. Enhanced data preprocessing and evaluation pipelines, raising accuracy from 87.5% to 100%, validated by human assessors. Collaborated with IT and cloud teams to publish and partially deploy a beta version of the chatbot to internal staff. Presented and explained chatbot operations to non-technical users.

System Engineer at Zephalto, Toulouse

Conducted R&D and prototyping of semi-automatic electronic systems for monitoring, adjusting, and optimizing oxygen concentration using real-time sensors. Designed electronic circuits, printed circuit boards, and system architecture, modelling with KiCAD. Performed in-depth analysis of system requirements, identification of technical constraints, and preparation of specifications. Managed manufacturing, integration, and testing of prototypes, including performance validation and implementation of optimization processes.

Research Assistant at LAAS-CNRS, Toulouse

Examined the impact of algorithms on machine learning accuracy, optimization solutions, and comparison of learning methods in federated architectures. Replicated federated learning systems as described in research papers. Optimized classification and algorithm accuracy.

Project Collaborator at Autonomous Vehicle Project

Collaborated with four teammates to prototype an autonomous vehicle capable of recognizing and reacting to road conditions using C++. Responsible for developing an embedded model able to analyze and classify real-time images using YOLOv5 technology. The group received a final grade of A+ for the project.

Project Contributor at Fil Rouge - Airbus

Worked with three teammates to develop a Python model for identifying rhetorical techniques used in memes. Developed a late fusion model combining two large language models to perform classification based on the textual content of the memes. The group received a final grade of A.

R&D and prototyping engineer at Zephalto

R&D and prototyping of semi-automatic electronic systems for monitoring, adjusting and optimizing oxygen concentration using real-time sensors. Design of electronic circuits, printed circuit boards and system architecture; modelling with KiCad. In-depth analysis of system requirements, identification of technical constraints and preparation of specifications. Manufacturing, integration and testing of prototypes, including performance validation and implementation of optimization processes.

Research Assistant at LAAS-CNRS, Toulouse

Embedded systems and AI tooling research: contributed to real-time data processing, AI model development, and hardware-software integration. Worked with Python, C++, SQL, and KiCad; performed system requirements analysis, specification drafting, prototype manufacture, integration, and testing to validate performance.

System Engineer at Téécom Paris

Involved in the integration of real-time systems, signal processing, and embedded operating systems. Worked on non-linear system analysis, data preprocessing, and software/hardware co-design; contributed to modeling with KiCad, data analysis, and prototype testing.

Projet Fil Rouge - Airbus: Memes Persuasion & Autonomous Vehicle Prototype at Airbus (Projet Fil Rouge)

Collaborated with a four-person team to (1) develop a Python model identifying rhetorical techniques in memes using a late-fusion approach that combines two large language models for textual-m content classification; and (2) prototype an embedded autonomous-vehicle perception model using C++ and YOLOv5 to analyze real-time road conditions and classify imagery. The project received a high final grade (A+).

R&D and prototyping of semi-automatic electronic systems for monitoring, adjusting and optimizing oxygen concentration using real-time sensors; design of electronic circuits, printed circuit boards and system architecture, modeling with KiCad; in-depth analysis of system requirements and technical constraints; preparation of specifications; fabrication, integration and testing of prototypes; validation and optimization of performance.