Moshe Gabel

About

Moshe Gabel’s research sits at the intersection of machine learning systems and computer architecture, with a particular focus on understanding and optimizing deep reinforcement learning (RL) workloads. His most cited work, “RL-Scope: Cross-Stack Profiling for Deep Reinforcement Learning Workloads” (2021, 4 citations), makes a foundational contribution by revealing that RL workloads exhibit fundamentally different computational and memory-access patterns compared to traditional deep learning—patterns that render standard profiling tools and system optimizations ineffective. By developing a cross-stack profiler that exposes these structural bottlenecks, Gabel provides the first systematic framework for diagnosing performance issues in RL systems, from the GPU kernel level up through the training loop. This work is especially impactful for practitioners deploying RL in robotics, data center management, and other latency-sensitive domains, where system-level inefficiencies can negate algorithmic advances. Gabel’s research is notable for bridging the gap between RL theory and practical system design, offering concrete guidance for building more efficient RL infrastructure—a critical need as RL workloads grow in scale and complexity.

Research Focus

Key Achievements

Top Papers

1. 1
   RL-Scope: Cross-Stack Profiling for Deep Reinforcement Learning Workloads

   4 citations · 2021

Key Collaborators

Contact & Links