About the RoleThe hardest problem in AI compute today isn't chips. It's power - permitted, interconnected, and available now. Base already has it, at thousands of sites, with more coming online every day.
Our answer to what that power is worth: a Base-built, Base-operated distributed GPU fleet that attaches datacenter-grade compute to our power network and serves it to the AI industry - from hosted hardware and bare-metal nodes up to an OpenAI-compatible inference API running on our own machines.
This is a founding-team role. You'll help build the whole thing: the control plane that inventories, activates, and recovers thousands of remote nodes; the provisioning path that flashes a server over Redfish/iDRAC from a thousand miles away; the agent that keeps every node trusted, observed, and working; and the dispatch layer that decides - together with our energy planning systems - when compute runs, throttles, or drains.
What You'll Do- Build the fleet control plane: inventory, identity, activation, telemetry, and operator tooling for a growing network of GPU nodes in the field
- Own out-of-band provisioning and recovery - Redfish/iDRAC automation, network boot, immutable OS images, and the rescue paths that make ad hoc SSH sessions unnecessary
- Develop the node agent: a secure, durable local substrate that establishes trust with the cloud, reports state, receives work, and survives reboots and bad networks
- Design overlay networking and secure connectivity across consumer internet links, and make it boring
- Connect compute to power: integrate dispatch with our energy planning systems so nodes run when power is available and cheap, and back off when the home or the grid needs it
- Work with hardware and deployment teams on enclosures, thermals, and the realities of running servers outdoors in a Texas summer
What You'll Bring- 8+ years building systems software close to hardware: platform management, firmware, provisioning, fleet orchestration, or backend services that operate physical machines
- Deep experience with server platforms - BMC/iDRAC, Redfish or IPMI, PXE/virtual-media boot, hardware inventory, and remote recovery
- Strong C, C++, Rust, or Go, and comfort moving between a boot log and a distributed service in the same day
- Experience operating fleets you couldn't walk up to: servers, network gear, vehicles, telecom, or energy hardware
- Judgment about what to build versus adopt - this project leans on industry standards and existing supply chains, not one-off inventions
- Ownership. Small team, new business line, real revenue targets. You'll set patterns others follow
Nice to Haves:- Datacenter or cloud platform background: hypervisors, bare-metal clouds, provisioning at scale
- GPU serving experience: vLLM, inference routing, KV-cache-aware scheduling, or GPU fleet operations
- Compiler, toolchain, or OS-image build depth
- Exposure to power systems or energy markets
You don't need energy experience. We'll teach you the grid. We can't teach 20 years of systems judgment.
About the TeamMost infrastructure jobs scale someone else's platform. This one starts a new kind of datacenter - one distributed across thousands of homes, powered by batteries we build, running compute we operate end to end. The problems are unsolved, the hardware is real, and the first nodes are going into the field now.
Please note: Base is a startup, which means priorities shift and evolve quickly. Your role may expand or change based on the needs of the business at any given time, so the responsibilities listed may not be exhaustive.
Please note: Base is a startup, which means priorities shift and evolve quickly. Your role may expand or change based on the needs of the business at any given time, so the responsibilities listed may not be exhaustive.
