Senior AI Training Architect — Distributed MoE Expert

NVIDIA; España; On-site; PHP 4,905,000 - 8,502,000; Full time

Job summary

NVIDIA seeks a senior Solutions Architect to drive large-scale distributed training and neural network alignment, supporting EMEA AI natives and research institutions in training and post-training workflows.

You will bridge research and engineering across the full training stack (including Nemotron) while shaping the NVIDIA product roadmap from customer feedback. Proficiency in Megatron-LM, NeMo, and PyTorch, plus experience with multi-node GPU clusters, is essential.

Qualifications

• MS or PhD in Computer Science, Engineering, or equivalent experience.
• 7+ years of distributed AI training experience with multi-node GPU clusters.
• Strong understanding of training infrastructure affecting efficiency and scalability.
• Proficiency with Megatron-LM and NeMo or equivalent frameworks.
• Excellent communication to engage researchers and engineers.

Responsibilities

• Build strategic technical relationships with leading AI model builders in EMEA.
• Define software stack and infrastructure for large-scale training and post-training workflows.
• Guide customers on distributed training strategies and RL-related workflows.
• Collaborate with product and research teams to convey customer needs.
• Support developer communities through demos, hackathons, and conferences.

Skills

Distributed AI training

Megatron-LM

NeMo

PyTorch

Communication skills

Multi-node GPU clusters

HPC

Education

MS or PhD in Computer Science or Engineering

Tools

Megatron-LM

NeMo

PyTorch

Job description

NVIDIA seeks a senior Solutions Architect to drive large-scale distributed training and neural network alignment, supporting EMEA AI natives and research institutions in training and post-training workflows.

You will bridge research and engineering across the full training stack (including Nemotron) while shaping the NVIDIA product roadmap from customer feedback. Proficiency in Megatron-LM, NeMo, and PyTorch, plus experience with multi-node GPU clusters, is essential.