ONE OF THE MOST IMPORTANT OUTSTANDING QUESTIONS IN EARTHQUAKE HAZARD RESEARCH IS WHETHER THE RUPTURE PROCESS IS DETERMINISTIC IN NATURE. THAT IS DO EARTHQUAKES KNOW HOW LARGE THEY LL GROW PRIOR TO ACHIEVING THEIR FINAL SIZE? IF SO WHEN WITHIN A MINUTES LONG RUPTURE PROCESS IS A VERY LARGE EARTHQUAKE DISTINGUISHABLE FROM AN ONLY LARGE ONE? THE EXISTENCE OF ANY DETERMINISM (OR LACK THEREOF) DEFINES THE MINIMUM POSSIBLE TIME AT WHICH CHARACTERIZATION OF AN EVENT AND ITS RESULTING HAZARDS CAN BE MADE. THUS THIS INFORMATION IS OF CRITICAL IMPORTANCE TO EARTHQUAKE AND TSUNAMI EARLY WARNING SYSTEMS. RAPID ACCURATE ESTIMATION OF MAGNITUDE AND GEOGRAPHIC EXTENT IS FUNDAMENTAL TO PROVIDING REAL-TIME ASSESSMENTS OF EXPECTED SHAKING AND TSUNAMI AMPLITUDES. WITHIN THE CONTINENTAL U.S. THE AFOREMENTIONED ISSUES ARE OF PRACTICAL IMPORTANCE. THE CASCADIA SUBDUCTION ZONE IS RECOGNIZED AS THE SINGLE GREATEST SOURCE OF SEISMIC AND TSUNAMI HAZARD IN THE PACIFIC NORTHWEST; GREAT MAGNITUDE 9 EARTHQUAKES HAVE OCCURRED ON IT AND WILL HAPPEN AGAIN. WHEN THIS DOES OCCUR GROUND SHAKING IN EXCESS OF 80%G (CONSIDERED VIOLENT TO EXTREME) WILL TAKE PLACE WITHIN THE FIRST 2 MINUTES OVER POTENTIALLY 1000 KM OF THE WESTERN PORTIONS OF CALIFORNIA OREGON AND WASHINGTON. WITHIN 5 MINUTES THE FIRST TSUNAMI WAVES WILL ARRIVE AT THE COASTLINE AND WILL GROW IN THE FIRST 15 MINUTES TO PEAK AMPLITUDES POTENTIALLY AS LARGE AS 20-30M. THE SEISMIC POTENTIAL IN CASCADIA WASN T FULLY REALIZED UNTIL THE LATE 1980S AND HENCE MUCH OF THE INFRASTRUCTURE IN THE REGION IS NOT BUILT TO WITHSTAND LARGE GROUND MOTIONS AND SUBSTANTIAL TSUNAMI INUNDATION. THE REGION IS FUNDAMENTALLY UNDERPREPARED COMPARED TO OTHER SUBDUCTION ZONES WORLDWIDE. FORTUNATELY A VAST NETWORK OF MORE THAN 600 REAL-TIME GNSS STATIONS EXISTS ACROSS THE WESTERN US TODAY WHICH IF PROPERLY HARNESSED CAN BE USED TO MAKE DETAILED REAL-TIME ASSESSMENTS OF THESE HAZARDS BEFORE THEY STRIKE THE POPULATION CENTERS. THIS PROPOSAL IS HYPOTHESIS-DRIVEN WE DEMONSTRATE THAT LARGE EARTHQUAKES ARE WEAKLY DETERMINISTIC AND RUPTURE PATTERNS CAN BE DISCERNED BY GNSS DATA THOUGH MACHINE LEARNING AND CAN BE LEVERAGED FOR ROBUST WARNING SYSTEMS IN ADVANCE OF HAZARDS. THE OBJECTIVE OF THIS PROPOSAL IS TO DEVELOP A FRAMEWORK UNDERPINNED BY HIGH-RATE GNSS OBSERVATIONS AND MACHINE LEARNING (ML) TO IDENTIFY THE FINAL MAGNITUDE EXTENT AND ASSOCIATED HAZARDS OF A LARGE EARTHQUAKE WELL BEFORE IT FINISHES RUPTURING.