Rack Design Engineer (Starmind)

RACK DESIGN ENGINEER (STARMIND)

The Starmind Compute team designs training and inference compute clusters from first principles, pushing thermal, electrical, and optical systems to the limit of physics. We operate as a highly cross disciplinary team comprised of mechanical, electrical, optical, and software engineers, all working to maximize the efficiency, reliability, and deployment speed of terrestrial computing to serve rapidly scaling demand.

This team owns compute architecture, datacenter network design, rack builds, connectivity, in-house design of our own computers, and the manufacturing line that builds that compute. As a Rack Design Engineer, you will own the rack as a product: structure, power, cooling, service paths, and the physical integration of servers, switches, optics, and liquid cooling from first sketch through production rate.

RESPONSIBILITIES:

• Own mechanical, structural, and integration design of high-density AI compute racks and associated chassis, frames, busbars, manifolds, and cable/fiber raceways
• Integrate compute, networking, power distribution, and liquid cooling into a buildable, serviceable, high-yield rack architecture
• Design for the in-house manufacturing line: DFM/DFA, fixtures, work instructions, incoming inspection, first article, and production ramp
• Close thermal, structural, and power constraints with cooling, electrical, optical, and network counterparts at the cluster level (CDUs, cold plates, quick disconnects, PDUs/busbars, high-current distribution, tray and bend-radius limits)
• Drive vendor and in-house parts from concept to qualified production, including sheet metal, extrusions, busbar, manifolds, cable management, and rack accessories
• Stand on the floor during prototype and production builds: debug fit, leak, cable dress, airflow/liquid balance, and service time; fold issues into the next revision the same day
• Produce and maintain CAD, BOMs, drawings, and build documentation that a line can run at volume
• Support cluster deployment of the racks you design, including first-site bring-up and feedback into the next hardware generation
• Partner with network, optical, electrical, and firmware teams to deploy custom compute for inference and training across SpaceX programs

BASIC QUALIFICATIONS:

• Bachelor’s degree in mechanical engineering, aerospace engineering, or another engineering discipline
• 2+ years of professional experience designing electromechanical hardware that was built and shipped (internship and project experience counts toward this if the hardware was physically produced)

PREFERRED SKILLS AND EXPERIENCE:

• Ownership of rack, chassis, or similarly dense electromechanical products through prototype and production
• Experience or background in ethernet switching, networking, interconnects, high power distribution, and other relevant data center technologies 
• CAD fluency in SolidWorks, Creo, or NX, with released drawings a shop and a factory can build from
• AI / HPC / GPU rack design: NVIDIA HGX/GB-class, custom OCP, or equivalent high-power liquid-cooled compute
• Liquid cooling design: cold plates, manifolds, CDUs, quick disconnects, leak detection, and coolant compatibility
• High-power rack electrical integration: busbar, PDU, 48V, high-current 3-phase, grounding, and service safety
• Cable and fiber physical design inside the rack (MPO/OSFP dress, bend radius, service loops) in partnership with optical and network engineers
• Experience taking a rack from NPI onto an in-house manufacturing line (fixtures, cycle time, poka-yoke, rework loops)
• Thermal and structural analysis (CFD, FEA) used to make a design decision
• Time at a high-velocity hardware organization: AI/GPU cloud, custom compute, or an OEM/ODM team that owned the product end to end
• Comfort in a factory or datacenter build environment, including first articles, yield, and hardware debug on the floor

ADDITIONAL REQUIREMENTS:

• Must be willing to work extended hours and weekends as needed to support design releases, line bring-up, and cluster deployments
• Must be willing to travel up to 30% to manufacturing and datacenter sites
• Ability to lift up to 25 lbs. unassisted and work around active factory and datacenter hardware