Ideaon
Website:
ideaoninc.com
Job details:
Location: Bangalore [Flexible]
Experience: 12 to 18 Years Exp
Type: Full-time
________________________________________
About Us
We’re building the next generation of AI SmartNICs — intelligent network accelerators designed to power distributed AI training, data center virtualization, and hyperscale cloud workloads.
Our mission: push networking performance to the limits — from microsecond-level latency to terabit-scale throughput — while giving developers full programmability across compute, network, and storage fabrics. If you love working close to the metal and want to shape the future of data-driven infrastructure, you’ll fit right in.
________________________________________
What You’ll Do
- Own and build NIC and RDMA kernel drivers for our SmartNIC platform (Linux).
- Design and implement high-performance Tx/Rx and RDMA data paths — QPs, CQs, doorbells, DMA, and WQEs.
- Bring up new silicon: work directly with hardware and firmware teams to validate PCIe, queue interfaces, and offload engines.
- Add networking offloads (TSO, checksum, RSS, VLAN, tunneling: VXLAN/Geneve/GRE).
- Integrate RoCEv2 and user-space RDMA (libibverbs/rdma-core) for AI and storage workloads.
- Build tools for performance measurement, kernel tracing, and packet analytics.
- Collaborate across firmware, DPDK, and P4/eBPF pipeline teams to build a unified data path.
- Push upstream when possible — we’re believers in open standards and open source.
________________________________________
What You’ll Bring
- 12+ years of Linux kernel or driver development experience (preferably network or RDMA subsystems).
- Strong in C programming, concurrency, and debugging at the kernel level.
- Solid understanding of:
NIC architecture — Tx/Rx rings, CQEs, doorbells, DMA engines.
RDMA fundamentals — QPs, PDs, MRs, WQEs, CQ handling.
PCIe, MMIO, and DMA interactions.
- Experience with rdma-core, DPDK, or low-latency networking stacks.
- Familiarity with SR-IOV, VFIO, and virtio.
- Passion for performance tuning and squeezing out every nanosecond.
________________________________________
Bonus Points
• You’ve worked with SmartNICs or DPUs (NVIDIA BlueField, Pensando, Broadcom Stingray, Marvell Octeon, etc.).
• You understand how AI/ML training clusters use RDMA for scale-out.
• Experience with eBPF/XDP, P4, or programmable data planes.
• You’ve contributed to Linux netdev or RDMA subsystems.
• You enjoy profiling, tracing, and reverse-engineering hardware behavior.
Click on Apply to know more.
