Rafael Pardinas

Rafael is a Staff Research Scientist/Engineer at ServiceNow AI Research, where he leads the reinforcement learning stack for the AI Research Group. His work spans both algorithms and training systems for reinforcement learning in LLM agents, centered on a core question: how can agents turn long-horizon experience into durable skills that generalise across tasks, domains, and environments? His current interests include long-horizon credit assignment in multi-turn RLVR, multi-domain training, memory-augmented agents that improve through open-ended interaction, and scalable post-training systems.   This dual focus on algorithms and infrastructure traces back to his earlier career as an engineer at Cisco, where he built distributed systems for enterprise communication. He later transitioned from machine learning engineering to applied research at Element AI, joining ServiceNow AI Research through its acquisition in January 2021. That engineering background continues to shape how he approaches RL research today, particularly in long-running training systems where algorithmic choices and infrastructure constraints are deeply connected.   Rafael holds a Master’s degree in Computer Security from the Universitat Oberta de Catalunya and studied Physics at the University of Santiago de Compostela.