Senior SRE AI Systems Performance

Luxoft; São Paulo; On-site; BRL 180,000 - 240,000; Full time

Job summary

Luxoft is seeking a software engineer focused on Distributed Inferencing on GPUs to improve key applications and benchmarks. You will join a core team driving performance on the latest hardware and software, tackling multi-GPU and multi-node scalability with emphasis on quality assurance and automation in AI/ML.

The role emphasizes collaboration with GPU library teams, development of performance dashboards, and ownership of impactful solutions across distributed systems, while maintaining robust

Qualifications

• Hands-on experience with AI inference or serving frameworks such as vLLM, SGLang, and Llama.cpp.
• Understanding KV cache transfer mechanisms and expert parallelization approaches (Mooncake, NIXL/RIXL, DeepEP, MORI, PPLX-Garden).
• Strong C/C++ and Python skills, with experience in software design, debugging, performance analysis, and test development.
• Experience running AI workloads on large-scale distributed systems.
• Familiarity with cluster management and orchestration platforms such as SLURM and Kubernetes.
• Experience with GitHub, Jenkins, or similar CI/CD tools and modern development workflows.

Responsibilities

• Distributed AI enablement and benchmarking on multi-GPU, multi-node clusters.
• Optimize AI workloads across scale-up, scale-out, and distributed configurations.
• Collaborate with GPU library teams to analyze distributed workloads for high throughput and low latency.
• Develop parallelization strategies to boost performance across diverse systems.
• Contribute to distributed model management systems, model zoos, and benchmarking pipelines.
• Build real-time dashboards reporting performance and accuracy for stakeholders.

Skills

GPGPU

C/C++

Python

AI Inference Frameworks

CI/CD

Distributed AI

Tools

SLURM

Kubernetes

GitHub

Jenkins

Job description

• A leading semiconductor and high-performance computing company is looking for a software engineer who is passionate about Distributed Inferencing on GPUs and improving the performance of key applications and benchmarks. You will be a member of a core team of incredibly talented industry specialists and will work with the very latest hardware and software technology

Responsibilities

• We are seeking a software engineer with strong technical expertise in C++/ Python development, solving performance and investigating scalability on multi-GPU, multi-node clusters. He is also passionate about quality assurance, benchmarking, and automation in the AI/ML space. The ideal candidate thrives in both collaborative and independent environments, demonstrates excellent problem-solving skills, and takes ownership in defining goals and delivering impactful solutions.
• KEY RESPONSIBILITIES:
• Distributed AI Enablement and Benchmarking: Enable and benchmark AI models on large-scale distributed systems to evaluate performance, accuracy, and scalability.
• Scalable Systems Optimization: Optimize AI workloads across scale-up (multi-GPU), scale-out (multi-node), and scale-across distributed system configurations.
• Cross-Team Collaboration: Collaborate closely with internal GPU library teams to analyze and optimize distributed workloads for high throughput and low latency.
• Parallelization Strategies: Develop and apply optimal parallelization strategies for AI workloads to achieve best-in-class performance across diverse system configurations.
• Model Infrastructure and Management: Contribute to distributed model management systems, model zoos, monitoring frameworks, benchmarking pipelines, and technical documentation.
• Performance Monitoring and Visualization: Build and maintain real-time dashboards reporting performance, accuracy, and reliability metrics for internal stakeholders and external users.

Mandatory Skills

• General-Purpose computing on GPUs (GPGPU)

Mandatory Skills Description

• PREFERRED EXPERIENCE:
• AI Framework Engineering: Hands-on experience with AI inference or serving frameworks such as vLLM, SGLang, and Llama.cpp.
• KV Cache and Expert Parallelization: Understanding KV cache transfer mechanisms and technologies (e.g., Mooncake, NIXL/RIXL) and expert parallelization approaches (e.g., DeepEP, MORI, PPLX-Garden).
• Programming and Software Design: Strong C/C++ and Python skills, with experience in software design, debugging, performance analysis, and test development.
• Large-Scale Distributed Systems: Experience running AI workloads on large-scale, heterogeneous compute clusters.
• Cluster and Orchestration Systems: Familiarity with cluster management and orchestration platforms such as SLURM and Kubernetes (K8s).
• Development Tools and Workflows: Experience with GitHub, Jenkins, or similar CI/CD tools and modern development workflows.