Alexandre Aguedo

Development and deployment of real-time AI applications using PyTorch, ONNX, FastAPI, and Docker. Projects focused on emotion recognition from facial video, speech audio, and text, all deployed on Hugging Face Spaces. Objective : build and conceive complete inference pipelines, from model training to real-world usable applications.

Feasibility study of a computer vision model to predict the outcome of the simulation stage in the treatment of mobile tumors (Cyberknife). Work conducted on medical imaging data: data collection, preprocessing, design and training of a suitable deep learning model, and performance evaluation. Results were presented to the teams at Accuray, the manufacturer of the Cyberknife.

Research project carried out during the first year of the Master's degree, focused on the automatic identification of verbal multiword expressions in French, as part of the PARSEME-FR initiative. Developed and evaluated NLP models for detecting fixed expressions in annotated corpora. Trained supervised models and analyzed results across different architectures.