Member of Technical Staff, Systems Infrastructure (2026 PhD New Grad)New$200K–$230K

Fireworks AI
Apply now

The opportunity

Fireworks is the platform for specialized intelligence, enabling companies to build, train, and serve AI models tailored to their own data, workflows, and products. Founded by the team behind PyTorch and backed by AMD, Atreides, Benchmark Capital, Index Ventures, Lightspeed,…

What you'll do

  • Scheduling & resource management: GPU job scheduling and scheduling of compute-intensive workloads across heterogeneous hardware; multi-tenant isolation, fair sharing and preemption, topology- and locality-aware placement, autoscaling, fleet utilization, and capacity planning across accelerator generations and vendors

  • Distributed storage & caching: High-performance distributed storage and caching for model weights, checkpoints, datasets, and KV cache; tiering across memory, local NVMe and object storage, cache admission and eviction policy, consistency, and fast cold-start and weight-loading paths

  • Datacenter networking: High-performance DC networks for AI workloads; RDMA/RoCE and InfiniBand, collective communication (NCCL/RCCL) performance, congestion control, topology design, load balancing, and tail-latency and reliability engineering at fleet scale

  • Design, build, and operate core infrastructure systems for large-scale training and inference

  • Model and measure system behavior: build the benchmarks, traces, and simulators needed to reason about scheduling, caching, and network performance before committing to a design

  • Identify bottlenecks across the stack, from kernel and driver to scheduler: policy, and drive them out with data

What they're looking for

  • Turn research ideas into production systems that hold up under real: workloads, real failures, and real customers
  • Work closely with the research and inference teams so that infrastructure: design and model design inform each other
  • Contribute to the team's technical direction by tracking emerging hardware,: interconnects, and systems research
  • PhD completed within the last 6 months, or expected completion by December: 2026, in Computer Science, Computer Engineering, Electrical Engineering, or a similar field