Gal Dalal

Papers

Total Citations

282

H-Index

About

Gal Dalal is a reinforcement learning researcher whose work sits at the critical intersection of safety, real-world deployment, and scalable systems. His most influential contribution, "Safe Exploration in Continuous Action Spaces" (2018, 275 citations), tackles one of the field's most pressing challenges: enabling RL agents to operate on physical systems — such as datacenter cooling units and robots — without ever violating safety-critical constraints. By exploiting the smooth dynamics typical of such systems, Dalal demonstrated that safe exploration is not only theoretically principled but practically achievable, a breakthrough that has resonated widely across the robotics and autonomous systems communities. Beyond safety, Dalal has pushed the boundaries of RL applicability by addressing the often-overlooked problem of action delays in real-world environments. His work on non-stationary Markov policies in delayed settings has meaningful implications for robotics, cloud computing, and finance. He has also contributed to the systems side of RL research, examining how CPU-GPU architectures can better support distributed training at scale. Together, these contributions paint the portrait of a researcher committed to bridging the gap between theoretical reinforcement learning and robust, deployable real-world systems.