Senior Distributed Systems Engineer

The Institute of Foundation Models (IFM) designs and operates ultra-scale GPU supercomputing systems to train next-generation foundation models. We believe performance, fault tolerance, and scalability are co-designed across model architecture, communication systems, runtime, and hardware topology.

This role sits at the core of that effort — driving communication performance, distributed reliability, and cross-layer optimization for large-scale training workloads.

We are looking for a deeply technical engineer to co-design and optimize the communication stack for large-scale distributed training, including hybrid parallelism and Mixture-of-Experts (MoE) workloads.

This is not a network operations role. This is a systems-level engineering position focused on performance engineering, distributed debugging, and communication-runtime co-design.

·Design and optimize expert-parallel and hybrid-parallel communication patterns

·Drive high-performance hierarchical collectives for MoE workloads

·Co-design runtime orchestration with communication topology awareness

·Reduce tail latency and improve determinism across thousands of GPUs

·Architect fault-tolerant distributed execution under real-world cluster failures

·Communication-compute overlap and topology-aware collective optimization

·Deep debugging of NCCL, RDMA, and custom communication layers

·Hybrid expert parallel strategies in modern large-scale MoE systems

·Elastic and resilient distributed job orchestration concepts

·Congestion analysis and routing optimization across InfiniBand/RoCE fabrics

·Microbenchmarking and performance modeling for communication-heavy workloads

·Hybrid expert parallel communication for Mixture-of-Experts training

·Scaling behavior under network pressure

·Distributed orchestration for elastic, large-scale training

·Fault detection and recovery in distributed GPU workloads

·Cross-layer bottlenecks: GPU ↔ NIC ↔ PCIe ↔ NVSwitch ↔ Fabric ↔ Scheduler

·Experience optimizing distributed training at 1,000+ GPU scale (or equivalent depth)

·Hands-on expertise with RDMA, InfiniBand, RoCE, and GPUDirect RDMA

·Deep familiarity with NCCL and/or UCX internals

·Strong systems programming ability (C/C++, Rust, or Go)

·Strong familiarity with modern model training frameworks such as PyTorch

·Ability to troubleshoot and profile training performance issues related to communication bottlenecks

·Ability to translate research ideas into production-grade optimizations

·Experience debugging distributed hangs, desynchronization, and performance regressions

·You can explain why an communication degrades at scale and how to fix it

·You have improved real cluster throughput via communication redesign

·You can trace a distributed hang across ranks and identify the root cause

·You are comfortable working at the boundary between hardware and runtime

·Include a link to your GitHub (required)

·Provide links to relevant distributed systems, HPC, or large-scale training projects

·Include a list of publications and/or public technical reports (if applicable)

·Describe the hardest distributed debugging problem you solved

·Include measurable performance improvements you have delivered

This position is eligible for visa sponsorship.

*Comprehensive medical, dental, and vision benefits 

 *Bonus

*401K Plan

*Generous paid time off, sick leave and holidays

*Paid Parental Leave

*Employee Assistance Program

*Life insurance and disability