Senior RDMA Performance Engineer

CodiLime is a software and network engineering industry expert and the first-choice service partner for global networking hardware and software providers as well as telecoms. The company operates since 2011 with 250+ employees and serves clients worldwide including the US, Japan, Israel, and Europe. The project involves SONiC validation on NVIDIA Spectrum-Based Platforms for AI Networking, operating within an Agile and startup-like environment.

• 7+ years of professional experience in network engineering, including 3+ years in data center network design and architecture.
• Hands-on experience with core network protocols and technologies such as BGP, EVPN/VXLAN, LACP, and ECMP.
• Deep understanding of AI-dedicated data center designs including Rail-Optimized Design (ROD) and Rail-Unified Design (RUD) topologies.
• Proven experience configuring and troubleshooting NVIDIA/Mellanox Spectrum switches and ConnectX NICs.
• In-depth knowledge of congestion control protocols for AI data centers, specifically RoCEv2, Priority-based Flow Control (PFC), and Explicit Congestion Notification (ECN).
• Ability to tune switch buffers and implement Quality of Service (QoS) policies.
• Experience with performance benchmarking and validation of RDMA-enabled AI data center fabrics.
• Proficiency in using AI agents/tools for automation, testing, and analysis.
• Strong technical documentation abilities and professional English communication skills.

Nice to have:

• Professional certifications such as CCNP/CCIE Enterprise/DC, JNCIP, or equivalents.
• Hands-on experience with network automation and validation frameworks (pyATS, NAPALM, Batfish, Nornir, NetBox).
• Exposure to CI/CD pipelines (GitLab CI, GitHub Actions, Jenkins) and Infrastructure as Code concepts.

• Configure, troubleshoot, and maintain NVIDIA/Mellanox Spectrum switches and ConnectX Network Interface Cards (NICs).
• Implement and tune advanced congestion control mechanisms essential for AI workloads, including RoCEv2, Priority-based Flow Control (PFC), and Explicit Congestion Notification (ECN).
• Optimize network environments to ensure maximum throughput, minimal latency, and minimal Job Completion Time (JCT) by performing setup tuning.
• Conduct extensive performance benchmarking and validation testing on AI-ready data center fabrics utilizing RDMA and the NVIDIA Collective Communication Library (NCCL).
• Leverage AI agents and specialized tools to automate routine network tasks, conduct rigorous testing, and analyze network telemetry.
• Author and maintain clear, high-quality technical documentation detailing network designs, configurations, and benchmarking results.