Software Engineer

Position Overview MangoBoost is engineering the full-stack AI infrastructure driving next-generation, hyper-scale AI data centers—spanning advanced AI networking, next-gen GPU systems, high-performance storage, and the full-stack software ecosystem. Within this mission, our team is building a high-performance network operating system based on SONiC, tailored for next-generation 800G switch platforms (e.g., Broadcom TH5). This network fabric is designed to tightly integrate with our custom DPUs, GPU servers, and storage appliances to deliver low-latency, high-throughput connectivity for AI workloads. As a Software Engineer on this team, you will build and maintain this custom SONiC distribution from an early stage, integrating the latest upstream releases and developing network features that bridge the switch fabric with end-host devices. We hire from junior to senior—scope and ownership scale with your experience. A note on languages: our production stack utilizes C/C++, Python, and Go, but we treat AI-assisted development as a first-class workflow. If you can ship correct, performant systems code with the help of modern AI coding agents, the specific language you've used before is not a barrier—your grasp of systems and networking fundamentals is what matters. Responsibilities & Opportunities
• Build, customize, and maintain a SONiC-based Network OS for 800G high-performance switches.
• Periodically integrate, port, and validate the latest open-source SONiC releases into our production environment.
• Develop and optimize data center network features via SAI (Switch Abstraction Interface), focusing on RoCEv2, congestion control (PFC, ECN), and hardware telemetry.
• Collaborate cross-functionally with DPU, GPU, and Storage teams to co-design and debug system-level interactions across the AI infrastructure stack.
• Profile network data paths, isolate performance bottlenecks, and optimize forwarding behavior for AI inference and training traffic.
• Write clear design documents and collaborate with a globally distributed, English-working team. Required Qualifications
• BS, MS, or PhD in CS, EE, CE, or equivalent practical experience (junior to senior welcome)
• Solid understanding of L2/L3 networking protocols (e.g., BGP, ECMP, VLAN, IPv4/IPv6) and data center topologies (e.g., Spine-Leaf)
• Working knowledge of Linux systems and networking fundamentals (kernel routing, Netlink, TCP/IP stack, debugging tools)
• Strong programming ability in at least one systems language (C/C++, Python, Go), and comfort picking up unfamiliar codebases—including with AI coding agents
• Ability to write technical documentation and communicate effectively in English Preferred Qualifications
• Hands-on experience with SONiC architecture and component development (e.g., SWSS, Orchagent, SyncD, Redis interactions)
• Familiarity with high-performance networking silicon (e.g., Broadcom Tomahawk series) and Switch Abstraction Interface (SAI)
• Experience with RDMA/RoCEv2 internals, lossless network tuning, and AI traffic characteristics
• Practical experience integrating network switches with PCIe-based end-host devices like DPUs or SmartNICs
• Open-source contributions to SONiC, SAI, or related networking infrastructure projects