26 Oct
|
jobmate.org
|
Bengaluru
26 Oct
jobmate.org
Bengaluru
Location – Bangalore / Pune / Kolkata / Indore
We are looking for a hands-on Networking Architect to design and validate networking for AI clusters: scale-out back-end fabrics, front-end and storage networks, and the software that runs them. You will lead customer architecture discussions, PoCs and presales, and build the validation that proves a fabric performs.
Key Responsibilities
- Architect AI back-end fabrics: RoCEv2 lossless Ethernet and InfiniBand, rail-optimized and leaf-spine topologies, congestion control (PFC, ECN, DCQCN).
- Design and validate SONiC-based fabrics, including fabric automation and telemetry.
- Build fabric validation and benchmarking: NCCL tests, RDMA perftest and job-level collective performance.
- Integrate NICs/DPUs and storage networks (NVMe-oF, GPUDirect RDMA).
- Track Ultra Ethernet and scale-up interconnects (NVLink, UALink) and turn them into offerings.
- Lead presales, PoCs and technical proposals.
Qualifications
Education
- Bachelor’s or master’s degree in computer engineering or computer science, Electronics, Electrical, (or equivalent experience).
Experience
- 12 to 16 years in networking product engineering or data center networking.
- At least 3 years on RDMA, HPC or AI fabrics.
Must-Have Technical Skills
- Data center networking: L2/L3, BGP/EVPN, leaf-spine.
- RDMA: RoCEv2 and/or InfiniBand, with PFC/ECN/DCQCN tuning.
- Linux networking stack and NIC configuration and troubleshooting.
- Hands-on fabric validation with traffic generators, NCCL tests and perftest.
- C /C++ or Go; Python is a plus.
Positive-to-Have Technical Skills
- SONiC/SAI development.
- NVIDIA Spectrum-X, Quantum InfiniBand, BlueField; Broadcom Tomahawk/Jericho.
- Ultra Ethernet specification; UALink/NVLink.
- Streaming telemetry (gNMI).
- Kubernetes networking for AI clusters (Multus, SR-IOV CNI).
📌 AI Fabric Networking Architect (Bengaluru)
🏢 jobmate.org
📍 Bengaluru