1.54M jobs liveBrowse

Principal Architect, AI Inference Networking - NIXL and Dynamo at Mellanox Technologies, Ltd. · Katori

Asks for

The ad says“Strong systems programming in C++ and Python, with performance critical code shipped.”
The ad says“Working knowledge of modern LLM serving, including prefill/decode disaggregation, KV cache management, tensor and pipeline parallelism, and at least one serving stack (Dynamo, TensorRT-LLM, vLLM, SGLang, Triton).”
The ad says“Fluent Mandarin and professional English, and comfortable working across China, Israel and US time zones.”
The ad says“Hands on experience with RDMA networking (InfiniBand or RoCE) and GPU memory movement.”
The ad says“Hands on experience with RDMA networking (InfiniBand or RoCE) and GPU memory movement.”
The ad says“Working knowledge of modern LLM serving, including prefill/decode disaggregation, KV cache management, tensor and pipeline parallelism, and at least one serving stack (Dynamo, TensorRT-LLM, vLLM, SGLang, Triton).”
The ad says“Working knowledge of modern LLM serving, including prefill/decode disaggregation, KV cache management, tensor and pipeline parallelism, and at least one serving stack (Dynamo, TensorRT-LLM, vLLM, SGLang, Triton).”
The ad says“Fluent Mandarin and professional English, and comfortable working across China, Israel and US time zones.”
The ad says“Working knowledge of modern LLM serving, including prefill/decode disaggregation, KV cache management, tensor and pipeline parallelism, and at least one serving stack (Dynamo, TensorRT-LLM, vLLM, SGLang, Triton).”
The ad says“Strong systems programming in C++ and Python, with performance critical code shipped.”
The ad says“Hands on experience with RDMA networking (InfiniBand or RoCE) and GPU memory movement.”
The ad says“Hands on experience with RDMA networking (InfiniBand or RoCE) and GPU memory movement.”
The ad says“Working knowledge of modern LLM serving, including prefill/decode disaggregation, KV cache management, tensor and pipeline parallelism, and at least one serving stack (Dynamo, TensorRT-LLM, vLLM, SGLang, Triton).”
The ad says“Working knowledge of modern LLM serving, including prefill/decode disaggregation, KV cache management, tensor and pipeline parallelism, and at least one serving stack (Dynamo, TensorRT-LLM, vLLM, SGLang, Triton).”
The ad says“Working knowledge of modern LLM serving, including prefill/decode disaggregation, KV cache management, tensor and pipeline parallelism, and at least one serving stack (Dynamo, TensorRT-LLM, vLLM, SGLang, Triton).”
The ad says“Working knowledge of modern LLM serving, including prefill/decode disaggregation, KV cache management, tensor and pipeline parallelism, and at least one serving stack (Dynamo, TensorRT-LLM, vLLM, SGLang, Triton).”

Inference has become a networking problem. With disaggregated serving, where prefill runs on one set of GPUs, decode on another, and KV cache streams between them, performance depends on how fast you can move tensors between GPUs, across nodes, and in and out of storage. Performance is only half of it. These deployments are elastic, so the data path needs dynamic APIs that let workers join and leave, connections form and tear down, and transfers reroute at runtime without stalling the pipeline. NIXL, NVIDIA's open source Inference Xfer Library, is the layer that solves this, and it sits underneath NVIDIA Dynamo and a growing list of inference engines.

China is where some of the world's most demanding inference deployments are being built. We are hiring a senior architect to work directly with the platform teams building them, understanding what breaks at scale and turning that into upstream code and roadmap. You will work daily with core NIXL and Dynamo engineers in Israel and the US.

What you'll be doing:

Drive NIXL adoption in China, from the first architecture conversation through prototype and benchmark to production rollout and upstream contribution.
Profile customer inference stacks and fix what's costing throughput and TTFT: KV cache movement, memory registration, transfer scheduling.
Write NIXL backends and plugins, extend Dynamo integrations, and contribute optimizations upstream to NIXL and GPUDirect class data paths.
Help define the APIs that make the data path dynamic, supporting elastic scaling, worker churn and runtime rerouting in production inference clusters.
Bring the region's requirements into NVIDIA's roadmap with enough technical authority to change it.
Grow the local community around NIXL through talks, benchmarks and reference architectures.

What we need to see:

M.Sc. or Ph.D. in CS/EE/CE, or equivalent depth earned in industry.
12+ years building or optimizing large scale distributed systems: AI inference or training infrastructure, communication libraries, high performance networking, or HPC runtimes.
Strong systems programming in C++ and Python, with performance critical code shipped.
Hands on experience with RDMA networking (InfiniBand or RoCE) and GPU memory movement.
Working knowledge of modern LLM serving, including prefill/decode disaggregation, KV cache management, tensor and pipeline parallelism, and at least one serving stack (Dynamo, TensorRT-LLM, vLLM, SGLang, Triton).
Strong communication skills, with the technical credibility to be taken seriously by customer engineering teams and by NVIDIA's core library maintainers.
Fluent Mandarin and professional English, and comfortable working across China, Israel and US time zones.

Ways to stand out from the crowd:

Open source contributions to NIXL, Dynamo, NCCL, vLLM, SGLang, TensorRT-LLM or Mooncake. We will read your commits.
Inference at genuine scale: thousands of GPUs, multi tenant clusters, strict SLOs.
CUDA, and comfort reading kernel and driver level code when the answer is down there.
A community you've built: a project people adopted, a talk people still cite.

You will help shape how NIXL is adopted across China, working alongside the engineers who design the transport layer itself. The work is public, and disaggregated inference is being defined right now.

NVIDIA is committed to fostering a diverse work environment and proud to be an equal opportunity employer. As we highly value diversity in our current and future employees, we do not discriminate (including in our hiring and promotion practices) on the basis of race, religion, color, national origin, gender, gender expression, sexual orientation, age, marital status, veteran status, disability status or any other characteristic protected by law.

Asks for

Values