This role focuses on building frameworks to enhance AI reasoning capabilities, developing distributed reinforcement learning systems, optimizing inference-time compute techniques like tree search and planning, and creating environments for agents. It requires experience with large-scale RL and distributed systems, and staying current with state-of-the-art algorithms.

The Sandbox service team at xAI builds and maintains a secure, scalable system that gives AI models safe, controlled access to computational environments. This infrastructure powers critical workloads across training and product, enabling models to run code, build software, interact with tools, and control applications. The role involves provisioning containers and virtual machines on large-scale clusters, granting models interactive control over these remote environments, and spans full-stack development from job orchestration and resource scheduling to filesystem performance tuning. It supports real-time code execution for products like Grok and reinforcement learning in training where models explore various tools.