This is a fully remote position, open to applicants in California.

📋 Description

• Create datasets and tailored models and algorithms to assess and benchmark models and end-to-end systems in areas such as LLM Security, Agentic Safety, Content Safety, Hallucinations, and ML Fairness.

• Generate datasets and methodologies for filtering training data and producing training datasets, including reinforcement learning environments and teacher models, across essential safety tracks.

• Investigate and implement model- and system-level strategies beyond post-training, such as Instruction Hierarchy or Risk Detection.

• Collaborate with engineers, data scientists, and research teams throughout NVIDIA to enhance AI safety solutions.

• Emphasize the measurement and enhancement of security, content safety, and inclusivity in cutting-edge multimodal LLMs, with a focus on advanced agentic safety.

⛳️ Requirements

• Master’s or PhD in Computer Science, Electrical Engineering, or a related quantitative discipline (or equivalent experience).

• Over 8 years of demonstrated experience in systems software engineering or machine learning engineering.

• At least 4 years of hands-on experience in post-training for LLMs, which includes SFT, RLHF/RLAIF, safety data generation methods, ablation studies, and deploying models into production.

• A minimum of 1 year of focused experience or research in LLM Security, Frontier Risks, Agentic Safety, or Multi-turn Safety Evaluation.

• Extensive understanding of machine learning principles and frameworks, ideally PyTorch.

• Strong proficiency in Python programming.

• Experience with large multimodal datasets and multi-modal foundational models.

• Exceptional analytical problem-solving skills along with excellent collaboration and communication capabilities.

• Traits that foster trust: humility, transparency, respect, and intellectual honesty.