Machine Learning Engineer, AI Safety

NVIDIA•Santa Clara, CA

1d•$124,000 - $241,500

About The Position

NVIDIA is developing AI-based products across multiple domains and collaborates with many AI companies. Ensuring the highest Content Safety possible reduces exposure to inappropriate material. Preventing Bias and Discrimination is essential to protect individual rights and achieve the best quality of results, including accuracy and completeness of information. By prioritizing safety and fairness, we can ensure that LLMs benefit everyone and contribute to a better future for all. Our team also works in the area of safety for generative models for language, robustness, and explainability. Our LLMs are a growing area of AI products, including models and services, and we are committed to ensuring that they are used safely and responsibly. We are looking for a talented Machine Learning Engineer to work on Product Security, Content Safety, ML Fairness and Robustness efforts for LLMs across all of our research and production engineering teams. In this role, you’ll have the opportunity to take on innovative problems in machine learning, particularly focused on safety for multi-modal LLMs. This role is directed at assessing, quantifying, and improving the safety and inclusivity of our LLM models in a scalable fashion.

Requirements

Master’s or PhD in Computer Science, Electrical Engineering or related field - or equivalent experience.
Minimum of 2+ years of work experience in developing and deploying machine learning models in production.
Strong understanding of machine learning principles and algorithms.
Hands-on programming experience in python and in-depth knowledge of machine learning frameworks, like Keras or PyTorch.
Background in one or more of the following broader areas for 1+ years: Content Safety, ML Fairness, Robustness, AI Model Security, or related areas.
Experience working in a range of the following areas within Content Safety: Hate/Harassment, Sexualized, Harmful/Violent, or other specific areas from your application.
Practice working with large multi-modal datasets and multi-modal models.
Good at problem-solving and analytical ability.
Excellent collaboration and communication skills.
Demonstrates behaviors that build trust: humility, transparency, respect, and intellectual honesty.

Nice To Haves

Skilled with alignment/fine-tuning of LLMs - including regular LLMs as well as VLMs (vision-language models) or any-to-text
Proven experience with multimodal and/or multilingual content safety, legal, and regulatory compliance.
Knowledge of robustness, including hallucinations, digressions, and generative misinformation.
Experience with GenAI security, including prompt stability, model extraction, confidentiality/data extraction, integrity, availability, and adversarial robustness.
Passion for AI and a demonstrated commitment to advancing the field through innovative research, prior scientific research, and publication experience.

Responsibilities

Develop the datasets and models for training and evaluating models and end-to-end systems for Content Safety, ProdSec, Robustness and ML Fairness.
Research and implement cutting-edge techniques for bias detection and mitigation in LLMs and systems using LLMs like RAGs.
Define and track key metrics for responsible LLM behavior and usage.
Follow the best MLOps practices of automation, monitoring, scale and safety.
Contribute to the MLOps platform and develop safety tools to help ML teams be more effective.
Collaborate with other engineers, data scientists, and researchers to develop and implement solutions to content safety and ML fairness challenges.