Staff Software Engineer, LLM Inference Platform

Zenlayer, Inc

United States

À distance

USD 180 000 - 240 000

Plein temps

14 jours+

Une candidature complète en une minute

Un CV et une lettre de motivation personnalisés, prêts à envoyer.

Passez les filtres ATS

Résumé du poste

Zenlayer, Inc. is seeking a Technical Lead for inference performance to own end-to-end Token Performance. You will reduce cost per million tokens and increase per-GPU goodput while meeting SLOs across models from billions to trillions of parameters.

The role requires deep knowledge of Transformer inference, kernel-level optimization, and orchestration boundaries. You will drive disaggregation, explore MoE parallelism, and optimize CUDA kernels across GPUs.

Qualifications

  • Deep understanding of Transformer inference compute vs memory bottlenecks.
  • Experience identifying and fixing bottlenecks in scheduling, kernel selection, or KV management.
  • CUDA kernel tuning and memory optimization with hands-on experience.
  • Bachelor’s degree in CS/EE; 6+ years in systems/perf engineering.

Responsabilités

  • Establish performance metrics and load-testing for token performance and throughput.
  • Tune kernel-level inference engines and optimize scheduling, prefill, and KV paths.
  • Define boundaries between inference engine and orchestration layers (Dynamo/llm-d).
  • Lead KV cache transfer disaggregation and autoscaling strategies.
  • Experiment MoE parallelism and wide-EP design; quantify trade-offs.

Connaissances

Transformer inference knowledge
CUDA kernel tuning
Nsight profiling
Performance engineering
GPU performance work

Formation

Bachelor’s degree in CS/EE or related field

Outils

Nsight
TensorRT-LLM
vLLM
SGLang
Dynamo llm-d

Description du poste

About Zenlayer

Zenlayer is the first massively distributed edge cloud provider, operating 290+ Points of Presence across 70+ countries. With accolades including PTC "Best Networking Innovation", "Outstanding Cloud, Data Center, or Interconnection Company", and Inc. 5000, we enable businesses to deploy and run applications at the edge, thus reducing network latency to the end user. Zenlayer focuses on improving connections worldwide, all while nurturing a unique company culture that values service, growth, and trust. As a "Zenplayer", you’ll be joining a community of success-driven team players that share the dream of powering a better-connected world.


About the Role

As the technical lead for inference performance, you will own end-to-end responsibility for the platform's Token Performance. For models ranging from hundreds of billions to trillions of parameters (including MoE), you will continuously reduce cost per million tokens and increase per-GPU Goodput while meeting TTFT/TPOT SLOs, and prove the gain from every optimization with reproducible data.


What You Will Work On


  • End-to-End Performance Attribution and Benchmarking Establish a unified Token Performance metrics framework (TTFT, TPOT, Goodput, GPU utilization, per-token cost), and build load-testing benchmarks that reflect real production traffic (high concurrency, long context, multi-turn Agent workloads) Perform bottleneck attribution for models such as DeepSeek, Kimi, GLM, Qwen, and MiniMax across compute, memory bandwidth, communication, scheduling, and queuing, and let the data determine which layer is the dominant constraint

  • Kernel-Level Inference Engine Optimization Dive deep into the source code of vLLM / SGLang / TensorRT-LLM, customizing and contributing upstream on critical paths such as the scheduler, Continuous Batching, Chunked Prefill, CUDA Graph, and KV block management Clearly define the responsibility boundary between the inference engine and orchestration layers such as Dynamo / llm-d, and reduce tail latency through request routing and load balancing

  • PD Disaggregation and KV Cache Transfer Lead the core design of Prefill-Decode disaggregation: the KV transfer path (NIXL/RDMA/UCX), transfer-compute overlap, and phase-aware autoscaling Define the failure domains and recovery semantics of cross-instance KV transfer, and answer the key question: under which workloads does PD disaggregation deliver gains, and under which does it make things slower

  • MoE Parallelism and Wide-EP Design hybrid TP/PP/DP/EP parallelism strategies and quantify the trade-offs between EP and TP Tackle expert load imbalance and All-to-All communication overhead, and drive Wide-EP into production

  • Kernels and Quantization Write and tune critical CUDA (and ROCm/HIP) operators: memory access optimization, kernel fusion, and Attention / MoE / communication-related kernels Build a closed-loop "quantization – accuracy – throughput" evaluation: the real-world gains and accuracy regressions of FP8/FP4/INT4/AWQ/GPTQ across different engines and hardware


Requirements

First-principles understanding of Transformer inference: the compute vs. memory-bound nature of Prefill and Decode, the KV Cache memory model, the differences among MHA/GQA/MLA, and how MoE routing constrains parallelism Source-level, in-depth experience with at least one mainstream inference engine, with real cases of identifying and fixing bottlenecks in scheduling, kernel selection, or KV management CUDA kernel tuning skills (Nsight profiling, memory optimization, fusion, CUDA Graph) with hands-on practical experience Bachelor’s degree in Computer Science, Electrical/Computer Engineering, or related field (HPC systems, ML Systems). Minimum 6 years in systems/Performance Engineering, with at least 2-3 years directly on LLM or GPU performance work Preferred, but not required: Experience with Merged PRs in vLLM Experience operating at cluster sale (hundreds to thousands of GPUs)

Obtenez votre examen gratuit et confidentiel de votre CV.

ou faites glisser et déposez votre fichier ici.

Similar jobs

Postes similaires à comparer

Machine Learning Engineer, LLM Inference Optimization
Machine Learning Engineer, LLM Inference Optimization

GMI Cloud, Inc • San Francisco (CA)

Sur place
USD 180 000 - 280 000
Member of Technical Staff, ML Systems ML Systems
Member of Technical Staff, ML Systems ML Systems

TensorScale • San Francisco (CA)

Sur place
USD 180 000 - 260 000
Staff Software Engineer — LLM Inference Platform Lead
Staff Software Engineer — LLM Inference Platform Lead

Zenlayer, Inc • États-Unis

À distance
USD 180 000 - 240 000
Machine Learning Engineer- Inference Optimization | Experienced Hire
Machine Learning Engineer- Inference Optimization | Experienced Hire

Susquehanna International Group, LLP • Bala Cynwyd (PA)

Sur place
USD 150 000 - 230 000
ML Systems Engineer — Performance & Scale
ML Systems Engineer — Performance & Scale

Veeda Innovation • California

Hybride
USD 150 000 - 230 000
Member of Technical Staff (AI Inference Engineer)
Member of Technical Staff (AI Inference Engineer)

FDE Times • San Francisco (CA)

Sur place
USD 140 000 - 240 000
Member of Technical Staff (AI Inference Engineer)
Member of Technical Staff (AI Inference Engineer)

Perplexity • San Francisco (CA)

Sur place
USD 140 000 - 210 000
AI Inference Kernel Optimization Engineer
AI Inference Kernel Optimization Engineer

Vidman AI • États-Unis

À distance
USD 140 000 - 210 000
Member of Technical Staff (AI Inference Engineer)
Member of Technical Staff (AI Inference Engineer)

Perplexity • New York (NY)

Sur place
USD 220 000 - 485 000
Member of Technical Staff (AI Inference Engineer)
Member of Technical Staff (AI Inference Engineer)

Perplexity • Palo Alto (CA)

Sur place
USD 220 000 - 485 000