The AI research team’s model training infrastructure has been upgraded with additional GPU capacity and improved data pipeline throughput. Training jobs can now be scheduled through the updated submission portal. Resource allocation policies have been documented.