Software Engineer – AI/HPC Middleware
Cornelis Networks, Inc.2 months ago
Remote, United StatesMid Level
Responsibilities
- Implement and optimize HPC middleware including MPI and SHMEM and AI collective communication stacks such as NCCL and RCCL.
- Integrate middleware capabilities with transports and provider layers including libfabric/OFI, UCX, and applicable verbs-like interfaces.
- Prepare and submit upstream contributions across MPI, SHMEM, collective communication, and related open-source projects.
- Collaborate with kernel, driver, and switch teams to validate end-to-end performance against the product roadmap.
- Analyze performance traces, reproduce customer issues, and help translate findings into fixes.
Requirements
- At least two years of systems programming experience in C/C++ on Linux.
- Bachelor’s degree in Computer Science, Engineering, or a related field, or equivalent experience.
- Exposure to or coursework/project experience with MPI, SHMEM, NCCL, RCCL, CUDA, ROCm, or other collective communication stacks.
- Familiarity with performance concepts and profiling or tracing tools, or strong willingness to learn.
- Foundational understanding of networking and/or RDMA concepts.
- Preferred qualifications include open-source contributions, libfabric provider experience, Ultra Ethernet specifications, RoCEv2, congestion control, Ethernet-based RDMA, benchmarking, profiling, optimization, Omni-Path/OPX, or Ethernet-based HPC fabrics.
Benefits
- Remote position for employees residing within the United States.
- Compensation package includes equity, cash, incentives, and potential performance-based bonuses.
- Medical, dental, vision, disability, life, dependent-care flexible spending, accidental injury, and pet insurance are available.
- Paid holidays, 401(k) with company match, Open Time Off for regular full-time exempt employees, sick time, bonding leave, and pregnancy disability leave are provided.
About Cornelis Networks, Inc.
Cornelis Networks builds high-performance, lossless scale-out interconnects for AI and HPC data centers, combining network adapters, switches, and fabric management software. Its products, based on the Omni-Path architecture, aim to accelerate GPU-, CPU-, and accelerator-based clusters for enterprises and research institutions. Founded in 2020 and headquartered in Wayne, PA, the company is privately held and sells hardware, software, and support to operators of large compute infrastructures.