Senior Machine Learning Engineer, AI Safety Evaluations

NVIDIA has helped redefine what computers can do. Now, generative AI is opening new possibilities for how people learn, create, and tackle problems. Realizing that potential means answering some of the hardest questions in AI: Where do models fall short? Which risks are we missing? How do we know our safeguards are working? As an expert in machine learning for Evaluations, you’ll help establish the evidence that guides the next generation of safer AI. Your curiosity and scientific judgment can shape what we test, what we learn, and what we build next. We are looking for a Machine Learning Engineer / AI Researcher / Data Scientist to join our AI Safety and Responsibility efforts within Enterprise Risk Management. Engaging directly in engineering tasks, you will apply your data science and statistics skills. You will turn open research questions into rigorous benchmarks, inventive experiments, and tools that reveal how multimodal large language models behave under pressure. Working with researchers and platform engineers, you’ll connect those discoveries to better training methods and safeguards, with reproducible results that inform technical reviews and executive decisions. If you love asking the question no one has thought to test, and building a convincing way to answer it, come help us define what progress in AI safety looks like. What You Will Be Doing: • Safety Benchmarking: Design and maintain safety benchmarks for large language and multimodal models, translating risk scenarios into test cases, datasets, and scoring rubrics. • Evaluation Methodology: Define metrics and protocols that assess safety alongside model usefulness, using baselines, uncertainty estimates, and failure analysis to distinguish meaningful improvements from noise. • Reproducible Evaluation: Build reusable Python pipelines with versioned datasets, model configurations, and experiment tracking to reproduce results and compare performance across model releases. • Automated Red-Teaming: Develop tools to discover failure modes and expand evaluation coverage, incorporating human review to validate automated scoring and provide additional judgment. • Failure Analysis: Investigate model failures with researchers, document actionable findings, and develop regression tests to assess whether model changes resolve identified issues. • Results and Communication: Partner with platform engineers to store and aggregate evaluation results, and clearly communicate findings, limitations, and emerging risks to researchers and decision-makers. What We Need To See: • Education and Experience: Bachelor’s degree or equivalent experience, with 6+ years in the field of machine learning engineering or evaluation. • Evaluation Design: Experience developing benchmarks, metrics, and statistically sound comparisons for ML systems, including selecting baselines and interpreting uncertainty.  Ability to translate ambiguous evaluation questions into testable experiments. • Engineering: Strong Python skills and experience building maintainable shared tools, data pipelines, and experiment-tracking workflows. • Analytical Rigor: Ability to assess dataset quality, identify limitations in metrics and scoring methods, and support conclusions with reproducible evidence. • Communication: Clearly explain findings and tradeoffs to researchers, engineering partners, and executives. Ways to Stand Out from the Crowd: • Evaluation: Experience evaluating LLMs, multimodal models, or agentic systems through adversarial testing or automated red-teaming. • Community Contributions: Contributions to public benchmarks or evaluation frameworks. • Research Publications: A publication track record and interest in continuing to publish at top tier venues. • Human-Aligned Evaluation: Experience