Network CCL Engineer
- Hiring from
- India, United Arab Emirates
- Work type
- Hybrid
- Posted
625,805 remote jobs, straight from company career pages
100% free · New jobs every minute
Show job descriptionHide job description
About us
We design custom server hardware for AI inference at datacenter scale. Our chips are purpose-built for running large AI models and are deployed in rack-level systems, where many devices work together to deliver low latency, high throughput, and energy efficiency.
About the role
We’re looking for an engineer to build the communication layer that lets our chips, and the GPUs alongside them, share data quickly while serving large AI models. You’ll develop our collective communication library, our equivalent of NCCL, from the low-level runtime and transport integration up to performance tuning at rack scale. You’ll work closely with our hardware, runtime, compiler, and model-serving teams to shape how our systems scale.
This is a low-level systems role focused on accelerator communication and distributed inference performance. It is not a web backend, general networking, or GenAI application role.
Requirements
- 5+ years of experience and an M.S. or Ph.D. in CS/CE (or equivalent experience)
- Strong C/C++ and Python, with experience in low-level systems: runtimes, drivers, networking, or distributed systems
- Solid knowledge of collective communication operations (AllReduce, AllGather, AllToAll, etc.) and libraries like NCCL, MPI, or UCX
- Understanding of GPU/accelerator memory and device-to-device data transfer
- Comfort working close to hardware: DMA, streams/queues, synchronization, and completion signaling
- Familiarity with high-speed networking such as InfiniBand, RoCE, or RDMA
- Strong performance tuning skills for bandwidth, latency, and overlapping compute with communication
Nice to Have
-
PyTorch distributed experience, including custom backends
-
Experience optimizing LLM inference communication, such as tensor parallelism, mixture-of-experts, or KV-cache transfer
-
Experience with serving frameworks like vLLM or SGLang
What We're Not Looking For
This role is not focused on general backend networking, web services, prompt engineering, or application-layer GenAI development. Instead, it is centered on low-level collective communication, accelerator runtime, heterogeneous device communication, and distributed AI performance at scale.
Why Join Us?
-
Build core infrastructure for next-generation AI inference hardware from the ground up
-
Solve hard problems spanning accelerators, networking, and distributed systems
-
Work across the full stack, from hardware to model serving
-
Directly shape the performance and scalability of our platform
