At Biofy, I developed and calibrated a fall-risk screening model and compared LSTM models for fall-risk activities. The work raised fall F1 from 79.7% to 82.1% and macro F1 from 71.4% to 75.8%.
At FILKOM UB, I developed and compared LSTM-based fall-detection models using skeleton data from YOLOv11, MediaPipe, and PoseFormer. I also built a reproducible experimentation pipeline for training, evaluation, and model comparison.
On my projects, I analyzed employee attrition, contributed to an automated micro-part counting system, and developed an RNA structure prediction pipeline that ranked 327th out of 1,877 teams in the Stanford RNA 3D Folding Kaggle competition.

