Senior Software Engineer, Compute Platform

About the role

Moonlite builds the infrastructure behind ambitious AI.

We design, deploy, and operate high-performance GPU systems for large-scale model training, computational research, and intensive data workloads. We combine bare-metal performance with the reliability and simplicity teams expect from the cloud.

What’s next in AI needs matching infrastructure. We’re building the team to make it happen.

Your Role:

You will be instrumental in building out our GPU-accelerated compute platform that powers distributed AI training and inference, large-scale simulations, and computational research workloads. Working closely with product, your platform team members, and infrastructure specialists, you’ll design and implement the compute orchestration layer that manages GPU clusters, bare-metal provisioning, and resource scheduling-enabling researchers and engineers to programmatically access high-performance compute resources with cloud-like simplicity.

Job Responsibilities

- Compute Orchestration Systems: Design and build scalable compute orchestration platforms that manage GPU clusters, bare-metal server provisioning, and resource allocation across co-located infrastructure environments.

- Resource Management & Scheduling: Implement intelligent workload scheduling, resource allocation, and optimization algorithms that maximize GPU utilization while maintaining performance guarantees for research and training workloads.

- Research Cluster Provisioning: Design and implement systems for provisioning and managing research computing environments including Kubernetes and SLURM clusters, enabling automated deployment, resource scheduling, and workload orchestration for distributed AI training and HPC workloads.

- GPU Platform Engineering: Develop platform capabilities for managing latest-generation NVIDIA GPU configurations (H100, H200, B200, B300), including GPU resource management, multi-tenant isolation, and integration with compute orchestration systems.

- Bare-Metal Lifecycle Management: Build automation and tooling for complete bare-metal server lifecycle management – from initial provisioning and configuration through ongoing operations, updates, and resource reallocation.

- Performance-Critical Systems: Optimize compute platform components for high-throughput and low-latency performance, ensuring research workloads achieve near-bare-metal efficiency in virtualized or containersized environments.

- Platform APIs & Integration: Develop robust APIs and SDKs that enable researchers to programmatically provision and manage compute resources, integrating seamlessly with existing workflows and research infrastructure.

- Observability & Monitoring: Implement comprehensive monitoring and telemetry systems for compute resources, providing visibility into GPU virtualization, workload performance and infrastructure health.

- Multi-Tenancy and Isolation: Build enterprise-grade multi-tenant compute isolation, security boundaries, and resource quotas that enable safe sharing of GPU infrastructure across teams and organizations.