GPU Platform Engineer

About the Role

LanceSoft is expanding its AI infrastructure practice and is looking for GPU Platform Engineers to design, build, operate and optimize GPU-accelerated platforms for enterprise clients. You will work on NVIDIA DGX and HGX based clusters, high-speed fabrics and AI workload platforms, and you will be the technical owner of platform health, performance and scalability from deployment through steady-state operations.

Required Certification

• NVIDIA-Certified Professional: AI Operations (NCP-AIO). Must be current and valid at the time of joining.

Preferred certifications

• NVIDIA-Certified Associate: AI Infrastructure & Operations (NCA-AIIO)
• Cloud certifications (AWS, Azure or GCP)
• CKA or RHCSA
• Terraform Associate

Key Responsibilities-

Platform design and deployment

• Design and deploy GPU clusters on NVIDIA DGX, HGX and OEM platforms (Dell, HPE, Supermicro and others), including compute, fabric and storage layers.
• Provision and manage clusters using NVIDIA Base Command Manager or equivalent tooling.
• Implement multi-tenant GPU platforms with workload isolation, quotas and fair-share scheduling.
• Support both on-prem and air-gapped deployments, including offline software repositories and container registries.

Orchestration and workload management

• Configure and operate Slurm and Kubernetes (including the NVIDIA GPU Operator, Network Operator and MIG configurations).
• Set up and maintain container workflows using Docker, containerd and NVIDIA NGC containers.
• Support AI and ML teams in onboarding training and inference workloads, and in right-sizing GPU allocation.

Networking and storage

• Configure and validate InfiniBand and RoCE fabrics, NVLink and NVSwitch topologies, and manage fabrics through UFM.
• Validate collective communication performance using NCCL tests and resolve bottlenecks.
• Integrate high-throughput storage for AI (parallel file systems, NFS, object storage) and tune data pipelines.

Performance, reliability and observability

• Benchmark and tune GPU, network and storage performance for training and inference workloads.
• Build observability with DCGM, Prometheus and Grafana, and define alerts, dashboards and SLOs.
• Diagnose GPU faults, XID errors, ECC and memory issues, thermal and power problems, and drive root cause analysis.
• Act as the escalation point for the NOC and manage vendor cases with NVIDIA and OEM support.

Automation and lifecycle management

• Automate provisioning, configuration and operations using Ansible, Terraform, Python and Bash.
• Own firmware, driver, CUDA and software stack compatibility and upgrade planning.
• Maintain reference architectures, runbooks, standards and platform documentation.
• Plan capacity and forecast GPU demand with clients and internal teams.

Client and team engagement

• Contribute technical input to client solution reviews, proposals and pre-sales discussions.
• Mentor L1 to L3 NOC engineers and share operational best practice.

Required Skills

• Deep understanding of the NVIDIA AI stack: DGX, Base Command, DCGM, NCCL, GPU Operator, NGC and NVIDIA AI Enterprise.
• Strong Linux administration, including kernel, driver and package management.
• Hands-on experience with Slurm and Kubernetes in GPU environments.
• Working knowledge of InfiniBand, RoCE and high-performance networking.
• Proficiency in scripting and infrastructure as code (Python, Bash, Ansible, Terraform).
• Solid grasp of monitoring and observability tooling and incident troubleshooting.
• Strong written and verbal communication, with the ability to explain platform decisions to technical and business stakeholders.

Qualifications

• Bachelor's degree in Computer Science, IT, Electronics or a related field (or equivalent experience).
• NCP-AIO certification, current and valid.
• Willingness to travel to client sites and to join on-call rotations for critical escalations where required.

Nice to Have

• Experience with NVIDIA Triton Inference Server, NeMo or other LLM serving and training stacks.
• Experience with MLOps platforms and GPU cost or utilization management.
• Experience with air-gapped or public sector AI environments.
• Familiarity with liquid cooling, power and data center design considerations for high-density GPU racks.
• ITIL 4 Foundation