This paper presents a machine learning-based Earthquake Early Warning System (EEWS) tested on the 28 March 2025 M7.7 Myanmar earthquake. Using ensemble learning algorithms with data from Myanmar Seismic Network station TGI, the system detected the earthquake within 3 seconds and reached M≥7.0 detection at 5 seconds post-P-arrival, providing 3.11 minutes warning before S-waves reached Bangkok. The system uses mel-frequency cepstral coefficients (MFCCs) and progressively estimates magnitude, location, and depth from expanding P-wave windows. Authors: Kasemsak Saetang, Wilaiwan Srisawat, Helmut Duerrast. Presented at TSER2025 Symposium, Bangkok, August 2025.