Juan Gonzalez

• Developed a Signal Quality Index algorithm (96% accuracy) reducing patient-monitoring false alarms across countries. • Built an inert-surface detection model (feature engineering + LDA, 98% accuracy) to prevent false vitals from devices off-patient • Led an XGBoost multimodal prototype combining physiological signals and qualitative data to forecast equipment needs and guide treatment planning • Designed a scalable data architecture for physiological data ingestion and ML training pipelines across studies

Grade: 4.8/5 • Member of CinfonIA research group. • Designed a U-shaped Nested Vision Transformer with deformable attention, reaching state-of-the-art 3D stroke segmentation. • Cut training time ~10% via CUDA optimization and SLURM scheduling on HPC. • Built reproducible training pipelines with PyTorch, MONAI, and Weights & Biases. • Deployed models in 3D Slicer via MONAI Label for clinical evaluation.

Grade: 4.3/5 Activities and societies: Member of PROMISE (Biomedical Signal and Image Processing) research groups. • Presented research at the XII International Research Meeting (UNITEC, 2020) and the 13th

Grade: 4.3/5 Activities and societies: Member of SemillIAS (Machine Learning Applied to Health) • Developed projects including EEG-based motion classification, atrial fibrillation detection from ECG, and MRI-based 3D reconstruction of knee cartilage. • Teaching Fellow in Biochemistry and Chemistry, mentoring students.