Chronic Chagas Disease is a parasitic cardiomyopathy often causing arrhythmias, conduction defects, and heart failure, yet early ECG diagnosis remains difficult, especially in low-resource. We propose \textit{SwissBeatsNet}, a multilead masked-autoencoder (MLMAE) framework that treats ECGs as synchronized channels to learn intra-lead temporal dynamics and inter-lead spatial dependencies. Self-supervised pretraining using CODE-15\%, SaMi-Trop, and PTB-XL datasets reconstructs randomly masked windows while incorporating an alignment objective to enhance representation learning. We then freeze the encoder and train an ensemble of linear heads to predict a Chagas disease score. On the hidden Physionet 2025 test set, the selected SwissBeatsNet model achieves a score of 0.425. Our team ranked 4th on the leaderboard.