Yassine Saoud

Developed a Python-based document search assistant using generative AI, achieving the following RAGAS metrics: 91% faithfulness, 79% answer relevancy, 74% answer correctness, and 97% context recall • Conducted a state-of-the-art review of existing technologies capable of addressing the problem • Implemented a modular RAG architecture integrating segmentation, retrieval (both dense and keyword-based), and generation (via OpenAI API and local models) • Tested the system by creating a 150-question dataset (both manually and automatically generated) • Optimized performance through iterative testing of retrieval methods and implementation of additional features: Query Translation, Semantic Chunking, and prompt optimization • Collaborated closely with domain experts to align the solution with their needs, present results, and incorporate feedback into development cycles

Developed a Python algorithm for waste reduction in industrial gate cutting, reducing wood waste by 500 kg per week from an initial 4000 kg (a 10% improvement), while also enhancing the algorithm’s optimization capacity • Decomposed the problem into logical sub-problems • Designed the solution using Constraint Programming (OR-Tools) and clustering algorithms (e.g. KMeans) • Independently managed the project lifecycle using Git, Poetry, and Pylint • Created a maintainable codebase with proper version control and comprehensive documentation

As part of a professional immersion internship, I was able to support teams of developers in a project, where I had the opportunity to do Python retro documentation