Responsibilities
- Own the NPI roadmap for GPU SKUs, including evaluation criteria, qualification timelines, and go-to-market strategy for new hardware generations
- Partner with datacenter teams to define requirements for power delivery (HVDC/LVDC), cooling (liquid vs. air), rack architecture, and physical infrastructure needed for next-gen GPUs
- Work with infrastructure engineers to validate hardware performance across key dimensions: training throughput (MFU), inference latency (TTFT, TBT), memory bandwidth, interconnect topology (NVLink, InfiniBand)
- Drive vendor engagement with NVIDIA, AMD, and emerging XPU providers—conducting technical deep dives, negotiating supply agreements, and managing early access programs
- Define product specifications for system configurations: single-GPU instances, multi-GPU nodes, full rack deployments, and megacluster topologies
- Analyze customer workload profiles to determine optimal GPU mix: H100 for large model training, L40S for inference, B200 for frontier research, MI300X for cost-sensitive workloads
- Build business cases for new SKU introductions, including CapEx requirements, depreciation models, utilization forecasts, and competitive pricing analysis
- Create technical documentation and benchmarking reports that help customers select the right GPU for their use case
- Monitor GPU availability, supply chain constraints, and allocation strategies to ensure Fluidstack can meet customer demand while maintaining healthy margins
- Collaborate with networking teams to ensure interconnect fabric (RoCE, InfiniBand) scales with GPU performance and supports distributed training patterns
Requirements
- 5+ years product management experience with at least 3 years focused on infrastructure, hardware platforms, or cloud compute services
- Strong technical background in GPU architecture, accelerator performance characteristics, and AI workload requirements
- Experience managing NPI processes from evaluation through production deployment—including vendor relationships, qualification testing, and rollout planning
- Deep understanding of datacenter infrastructure: power distribution, thermal management, rack design, and high-density deployment constraints
- Track record of making build vs. buy decisions on hardware platforms based on TCO analysis, competitive positioning, and customer demand signals
- Familiarity with GPU performance metrics (TFLOPS, HBM bandwidth, TDP, MFU) and how they translate to real-world training and inference performance
- Ability to work with engineering teams to debug hardware issues, analyze telemetry data, and identify root causes of performance degradation
- Experience conducting competitive analysis of cloud GPU offerings from AWS, GCP, Azure, CoreWeave, Lambda Labs, and other specialized providers
- Comfortable navigating supply chain complexity, allocation negotiations, and procurement timelines with hardware vendors
Nice-to-Haves
- Experience with networking topologies (fat tree, rail-optimized), storage systems (NVMe, Ceph), or HPC infrastructure design
Compensation
Cash compensation range: $150,000-$250,000 (base salary, equity, benefits; final offers vary based on geography, experience, credentials, and other factors)
FluidstackSan Francisco, CA
