Oumaima Ouayres

I build practical AI products from model experimentation through APIs, databases, interfaces, testing, and deployment. My work spans machine learning, NLP, generative AI, and full-stack software development.

For ApplyLens AI, I built an end-to-end application that analyses PDF and text documents to generate structured, evidence-grounded results. I developed its FastAPI, PostgreSQL, and pgvector backend alongside a React and TypeScript frontend, authentication, automated testing, and Docker deployment.

At LAPSSI at Cadi Ayyad University, I developed a character-aware XLM-RoBERTa model for multilingual research, created a 4,562-post dataset, and designed reproducible training and evaluation experiments. I used ablation studies and error analysis to identify the technical choices that improved the model.

I've also built scientific-text summarisation with T5 and BART at 3DSMARTfactory, a Java and React fleet-management platform at SALVIA, and research projects in speaker verification, generative modelling, and interpretable fraud detection.