Network Engineer, Design & Engineering

Fluidstack•Austin, TX

3d•$180,000 - $300,000

About The Position

Fluidstack is seeking a Network Engineer, Design & Engineering to join their Network Engineering team. This is a design-ownership role where the individual will take customer requirements (GPU shape, workload profile, scale targets, tenancy model) and produce end-to-end network architectures that are deployable, validated, and optimized for AI training and inference workloads. This role involves owning the full design problem space, including topology selection, rack layout implications, power and thermal constraints, cable plant feasibility, and fabric scaling. The engineer will work closely with cross-functional partners in Hardware, DC Operations, ICT/Structured Cabling, Software Engineering, and Validation to ensure designs are technically sound, physically buildable, and operationally sustainable. Success is defined by producing network designs that deployment teams can execute without ambiguity, scale to the customer’s target, and meet performance requirements on the first turn-up.

Requirements

5+ years of network engineering experience with a demonstrated focus on network design and architecture rather than purely operational roles.
Designed datacenter network fabrics from requirements through deployment — not just configured them.
Ability to articulate why a design decision was made, what tradeoffs were considered, and what constraints drove the outcome.
Strong command of datacenter network fundamentals including CLOS/fat-tree topologies, BGP (eBGP underlay, iBGP/eBGP overlay), EVPN/VXLAN, IP addressing and subnetting at scale, and physical layer design (optics selection, fiber types, link budgets).
Understand how L1 decisions cascade into L2/L3 behavior and design accordingly.
Working knowledge of RDMA network design (InfiniBand and/or RoCEv2), lossless Ethernet configuration (PFC, ECN, DCQCN), and the network performance requirements of distributed AI training workloads.
Understand why fabric design decisions directly impact training job completion time.
Experience designing networks around specific GPU platforms (NVIDIA DGX/HGX, AMD MI-series, custom accelerator platforms).
Understanding of how GPU topology, NVLink/NVSwitch architecture, and host networking configuration interact with fabric design.
Ability to reason about network design in the context of physical constraints. Experience working through rack layout planning, power budget allocation, structured cabling architecture, and equipment placement decisions.
Break complex design problems into fundamental components and reason through them systematically.
When faced with a new GPU platform, an unfamiliar facility constraint, or a novel customer requirement, decompose the problem rather than reaching for the nearest template.
Challenge assumptions — including your own — and can defend your design decisions with rigorous reasoning.
Produce design documentation that is clear, complete, and actionable. HLDs and LLDs enable deployment teams to execute without requiring you in the room.
See documentation as a design artifact, not an afterthought.
Excellent at working across engineering disciplines.
Communicate design intent clearly to non-network stakeholders (hardware, facilities, cabling) and incorporate their constraints into your designs.
Earn trust through technical depth and follow-through.

Nice To Haves

Experience designing networks at hyperscale companies (Meta, Google, Microsoft, AWS) or large AI infrastructure providers.
Experience seeing what disciplined design processes look like at scale and can adapt those patterns to a fast-growing startup.
Deep familiarity with multiple network hardware platforms (Arista, Juniper, NVIDIA/Mellanox, Broadcom-based).
Experience designing for specific platform capabilities and constraints, including ASIC-level considerations that impact fabric design.
Experience designing networks with automation in mind — consistent naming conventions, structured data models, templatable configurations.
Experience with WAN topology design, DCI (Data Center Interconnect), optical transport, and backbone network architecture.
Understanding of how campus/datacenter design connects to broader network infrastructure.
Experience building something from scratch before — ideally in a high-growth infrastructure or cloud company.
Comfortable with rapid context switching, evolving requirements, and the intensity of early-stage company building.

Responsibilities

Own the design lifecycle from customer requirements through deployable architecture. Produce topology designs, IP/addressing schemes, routing policy, and fabric configuration specifications for AI training and inference fabrics. Design front-end (out-of-band management, customer access), back-end (GPU-to-GPU training fabric), and storage network architectures.
Design network architectures that adapt to different GPU platforms (NVIDIA, AMD, custom accelerators), server form factors, and workload profiles. Each customer engagement may require a different rack layout, power envelope, cable infrastructure approach, and fabric topology.
Translate logical network designs into physical reality. Work cross functionally on rack elevation planning, power distribution constraints, structured cabling architecture (fiber trunk design, patch panel layouts, cable pathway routing), and cooling/airflow considerations that impact network equipment placement. Ensure designs are buildable within the physical constraints of each facility.
Produce comprehensive design packages that enable deployment teams to execute independently. This includes High-Level Designs (HLDs), Low-Level Designs (LLDs), cutsheet specifications, bill of materials, cabling matrices, and design decision records.
Design lossless Ethernet fabrics optimized for RDMA (RoCEv2) workloads including PFC configuration, ECN tuning, traffic class design, and congestion management. Understand the relationship between fabric topology, ECMP behavior, and collective communication patterns in distributed training workloads.
Partner with Hardware Engineering on server/GPU platform integration, DC Operations on facility constraints and power planning, ICT on structured cabling feasibility and fiber budgets, Software Engineering on automation requirements and DCIM data modeling, and Validation teams on test plans and acceptance criteria. Ensure designs satisfy constraints across all these domains.
Participate in and lead design review sessions. Contribute to the development of reference architectures, design standards, and reusable design patterns that accelerate future deployments. Challenge assumptions — both your own and others’ — to ensure designs are technically rigorous and operationally sound.