AI Fabric Networking Architect – AI Infrastructure

Sourcebae; Bengaluru; On-site; INR 4,000,000 - 7,000,000; Full time

Job summary

Sourcebae in Bengaluru seeks a hands-on AI Fabric Networking Architect to design, implement, and validate high-performance networking for AI clusters.

You will lead architecture discussions, PoCs, and technical proposals, with deep expertise in RoCEv2, InfiniBand, and RDMA. On-site role in India, focusing on scalable AI infrastructure and fabric validation.

Qualifications

• Bachelor's or master's degree in Computer Engineering, Computer Science, Electronics, Electrical Engineering, or equivalent practical experience.
• 12-16 years of experience in networking product engineering or data centre networking.
• At least 3 years of hands-on experience with RDMA, HPC, or AI networking fabrics.
• Strong understanding of data centre networking, including Layer 2/Layer 3, BGP/EVPN, and leaf-spine architectures.
• Practical experience with RoCEv2 and/or InfiniBand, including PFC, ECN, and DCQCN configuration and tuning.
• Strong knowledge of the Linux networking stack, NIC configuration, and network troubleshooting.
• Hands-on fabric validation experience using traffic generators, NCCL tests, and RDMA perftest.
• Programming experience in C, C++, or Go. Python is an added advantage.

Responsibilities

• Design AI backend fabrics using RoCEv2 lossless Ethernet and InfiniBand, including rail-optimized and leaf-spine topologies.
• Configure and optimize congestion-control mechanisms, including Priority Flow Control (PFC), Explicit Congestion Notification (ECN), and DCQCN.
• Design, implement, and validate SONiC-based networking fabrics, including automation, telemetry, and operational workflows.
• Develop and execute fabric validation and benchmarking using NCCL tests, RDMA perftest, traffic generators, and job-level collective performance measurements.
• Integrate NICs, DPUs, and storage networking technologies, including NVMe over Fabrics (NVMe-oF) and GPUDirect RDMA.
• Evaluate emerging technologies such as Ultra Ethernet, NVLink, and UALink and translate relevant capabilities into customer-ready infrastructure offerings.
• Lead customer architecture discussions, technical presales, PoCs, solution demonstrations, and technical proposal development.
• Collaborate with engineering and customer teams to troubleshoot performance bottlenecks, validate network behaviour, and optimize AI cluster connectivity.

Skills

RDMA

HPC

AI networking fabrics

RoCEv2

InfiniBand

PFC

ECN

DCQCN

NCCL tests

RDMA perftest

NVIDIA Spectrum-X

Education

Bachelor's/Master's in Computer Engineering/CS/EE

Tools

NCCL

RDMA perftest

SONiC/SAI

GPUDirect RDMA

Multus

SR-IOV CNI

Job description

Job Title: AI Fabric Networking Architect - AI Infrastructure

Experience: 12-16 Years

Work Mode: On-site

Job Summary

We are seeking a hands-on AI Fabric Networking Architect to design, implement, and validate high-performance networking solutions for AI clusters, including scale-out backend fabrics, frontend networks, and storage networks.

The ideal candidate will have deep expertise in data centre networking, RoCEv2 and/or InfiniBand, RDMA, congestion control, and large-scale fabric validation. The role also involves leading customer architecture discussions, technical presales, proofs of concept (PoCs), and technical proposals for AI infrastructure solutions.

Key Responsibilities

• Design AI backend fabrics using RoCEv2 lossless Ethernet and InfiniBand, including rail-optimized and leaf-spine topologies.


• Configure and optimize congestion-control mechanisms, including Priority Flow Control (PFC), Explicit Congestion Notification (ECN), and DCQCN.


• Design, implement, and validate SONiC-based networking fabrics, including automation, telemetry, and operational workflows.


• Develop and execute fabric validation and benchmarking using NCCL tests, RDMA perftest, traffic generators, and job-level collective performance measurements.


• Integrate NICs, DPUs, and storage networking technologies, including NVMe over Fabrics (NVMe-oF) and GPUDirect RDMA.


• Evaluate emerging technologies such as Ultra Ethernet, NVLink, and UALink and translate relevant capabilities into customer-ready infrastructure offerings.


• Lead customer architecture discussions, technical presales, PoCs, solution demonstrations, and technical proposal development.


• Collaborate with engineering and customer teams to troubleshoot performance bottlenecks, validate network behaviour, and optimize AI cluster connectivity.

Required Qualifications

• Bachelor's or master's degree in Computer Engineering, Computer Science, Electronics, Electrical Engineering, or equivalent practical experience.


• 12-16 years of experience in networking product engineering or data centre networking.


• At least 3 years of hands-on experience with RDMA, HPC, or AI networking fabrics.


• Strong understanding of data centre networking, including Layer 2/Layer 3, BGP/EVPN, and leaf-spine architectures.


• Practical experience with RoCEv2 and/or InfiniBand, including PFC, ECN, and DCQCN configuration and tuning.


• Strong knowledge of the Linux networking stack, NIC configuration, and network troubleshooting.


• Hands-on fabric validation experience using traffic generators, NCCL tests, and RDMA perftest.


• Programming experience in C, C++, or Go. Python is an added advantage.

Good-to-Have Skills

• SONiC/SAI development and networking fabric automation.


• NVIDIA Spectrum-X, Quantum InfiniBand, BlueField, and Broadcom Tomahawk/Jericho platforms.


• Ultra Ethernet specifications and scale-up interconnect technologies such as NVLink and UALink.


• Kubernetes networking for AI clusters, including Multus and SR-IOV CNI.