Solutions Architect, Pre-training and Post-training

NVIDIA NVIDIA · Semiconductors · Seoul, South Korea

NVIDIA is seeking a Solutions Architect to assist researchers and developers in accelerating their AI workloads using NVIDIA's platform. The role involves creating technical engagements, proposing state-of-the-art training and optimization frameworks, and promoting collaborative results. Requires 5+ years of experience in the full AI model lifecycle, including pre-training, fine-tuning, post-training, optimization, and evaluation, along with strong software engineering skills.

What you'd actually do

  1. Create fruitful technical engagements with AI development teams in frontier model makers in Korea and lead strategic relationships with top developers and influential researchers.
  2. Help them develop AI models more efficiently by proposing state-of-the-art training and optimization frameworks including Megatron-LM, Megatron-Bridge, NeMo-RL, NeMo-Gym, TensorRT Model Optimizer, and TensorRT-LLM.
  3. Promote the results of the collaboration between NVIDIA and those teams with the support of marketing teams by publishing press releases and celebrate together by presenting them at GTC.
  4. Continuously keep up with the latest AI training and optimization technologies that not only NVIDIA but also the community researchers provide to the market.

Skills

Required

  • 5+ years of hands-on experience in full AI model lifecycle, including pre-training, supervised fine-tuning, post-training such as reinforcement learning, optimization, and evaluation.
  • Strong software engineering skills, including debugging, performance analysis, and test development.
  • World-class communication skills with a demonstrated ability to articulate a value proposition to technical and non-technical audiences.
  • MS/PhD in Computer Science or Engineering or equivalent experience.

Nice to have

  • Excellent English communication skills
  • Understanding of infrastructure factors that can affect AI model development such as GPU architecture, server block diagram, or networking bandwidth among GPU servers or between GPU servers and shared storage.

What the JD emphasized

  • 5+ years of hands-on experience in full AI model lifecycle, including pre-training, supervised fine-tuning, post-training such as reinforcement learning, optimization, and evaluation.

Other signals

  • customer-facing technical expertise
  • deploying AI workloads at scale
  • pre-training and post-training optimization