Software Engineering, Compute Infrastructure

Dedalus Labs builds persistent computing environments for AI agents. The role owns GPU compute infrastructure end to end, from server deployment and fleet operations to customer-facing APIs and integration with the company’s virtual machine product.

Responsibilities:

• Bring GPU capacity online. Work with datacenter operators, hardware vendors, and onsite technicians to turn delivered servers into usable capacity. Plan and verify racking, power, cooling, cabling, networking, and remote management. Automate provisioning, operating-system images, firmware, drivers, and configuration. Establish acceptance tests that exercise GPUs, memory, storage, and network performance before machines enter the fleet
• Build the compute control plane. Design and implement the APIs, controllers, and services that allocate, provision, and manage GPU machines. Make resource state, quotas, permissions, and failure behavior clear to customers and other engineers. Reason about concurrency, durable state, repeated requests, cancellation, and partial failures. Measure provisioning time, API latency, and reliability under representative load
• Integrate GPU compute with Dedalus Machines. Work with our control plane and Rust virtual machine monitor, the software that runs each VM. Design how GPUs are assigned, isolated, exposed to guests, and safely reclaimed. Follow problems through Linux, device drivers, virtualization, storage, and networking. Build interfaces that let the GPU fleet and VM product evolve together
• Own production reliability. Build monitoring and diagnostics for machine health, GPU failures, network faults, and customer-visible service behavior. Define availability targets and the operating practices needed to meet customer SLAs. Automate detection, isolation, and recovery of unhealthy nodes. Coordinate repairs and replacements with vendors, lead incident investigations, and turn recurring failures into software fixes. Make upgrades, maintenance, and recovery safe and understandable
• Improve the systems around the fleet. Profile workloads and trace bottlenecks across compute, memory, networking, storage, and scheduling. Improve useful capacity and performance with measured changes. Choose existing tools where they fit, and build the parts our product needs. Establish tests, deployment workflows, documentation, and operating procedures that other engineers can use and maintain