Senior Forward Deployed Engineer, Open Models

NVIDIA•Us, CA

About The Position

NVIDIA is hiring a Senior Forward Deployed Engineer to drive technical engagements with partners and customers across its open model portfolio, including NVIDIA Nemotron and NVIDIA Cosmos. The role turns open models into production-ready AI customized for specific domains, from post-training through harness optimization and deployment. Work is primarily conducted through NVIDIA’s partner ecosystem, alongside partner engineering and research teams, with direct customer engagement when needed. The engineer guides AI solutions from discovery and scoping through post-training, harness optimization, and deployment. They develop use-case-specific strategies covering methods, data, evaluation, and success criteria, then create scalable architectures and deployment plans. Hands-on work spans evaluation design, synthetic data pipelines, parameter-efficient training such as LoRA, supervised fine-tuning, synchronous and asynchronous reinforcement learning, model architecture and hyperparameter choices, harness tuning, and inference optimization. The role leads technical engagements, shapes statements of work, and drives delivery with partner engineering teams and joint customers. It also creates reusable recipes, reference pipelines, and playbooks that enable partners to succeed independently. The engineer connects NVIDIA model research with partner and customer needs and shares field insights with NVIDIA product, research, and engineering teams to improve Nemotron, Cosmos, and the software stack.

Requirements

Bachelor’s, Master’s, or PhD in Computer Science, Engineering, Machine Learning, or a related field, or equivalent experience.
8+ years in ML or software engineering, applied AI, forward deployed engineering, or solutions architecture, including 3–5+ years of hands-on LLM post-training and taking models to production.
Solid fundamentals in deep learning and machine learning.
Deep expertise in post-training methods, including LoRA/PEFT, supervised fine-tuning, and reinforcement learning; understanding of synchronous and asynchronous RL trade-offs.
Strong grasp of model architectures and hyperparameters and their effects on training outcomes.
Good understanding of large-scale distributed training and inference performance optimization.
Ability to configure and tune workloads for memory efficiency, throughput, latency, and scaling across multi-GPU and multi-node systems.
Hands-on experience with post-training data, including synthetic data pipelines, dataset curation and filtering, and evaluation sets.
Experience with harness engineering for LLM and agentic systems.
Proficiency in PyTorch, Hugging Face Transformers, Python, and Linux.
Proven ability to scope engagements, shape statements of work, and lead delivery with partners and customers.
Excellent organizational skills, empathy for partner and customer needs, and ability to drive complex decisions across teams and companies through trusted relationships.
Ability to communicate complex technical ideas clearly to audiences ranging from engineers to C-level executives.
Strong problem-solving skills, intellectual curiosity, and commitment to partner and customer success.

Nice To Haves

Hands-on technical expertise and ability to describe solution paths for complex challenges.
History of enabling partners and customers to adopt open models and NVIDIA technologies and platforms successfully.
Technical vision that influences product and engineering decisions and drives production outcomes.
Hands-on experience with Nemotron, Cosmos, and other open models; serverless and self-managed post-training platforms (e.g. Tinker APIs); and open-source frameworks such as NVIDIA NeMo and Megatron.
Experience with multimodal or world-foundation-model post-training.
Experience running training jobs on large GPU clusters with SLURM or Kubernetes.

Responsibilities

Drive AI solutions from discovery and scoping through post-training, harness optimization, and deployment with partners and customers.
Develop post-training and harness strategies for each use case, including appropriate methods, data, evaluations, and success criteria.
Design evaluations and synthetic data pipelines; post-train models using LoRA and other parameter-efficient methods, supervised fine-tuning, and synchronous or asynchronous reinforcement learning.
Make model architecture and hyperparameter choices to improve training outcomes.
Tune LLM and agentic-system harnesses, including prompting, tool use, orchestration, and context management.
Optimize inference to support cost-effective production deployments.
Lead technical engagements by scoping work, shaping statements of work, and driving delivery with partner engineering teams and joint customers.
Build reusable recipes, reference pipelines, and playbooks that help partners succeed independently and scale domain-specific AI development.
Connect NVIDIA model research to partner and customer use cases and share field insights with NVIDIA product, research, and engineering teams.