Principal Research Scientist, Synthetic Data Generation

NVIDIA is at the forefront of the AI revolution, and our research is shaping the future of large language models. We are looking for a Principal Scientist to set the technical direction for synthetic data generation across NVIDIA's frontier model efforts. You will define and build open-source libraries within the NVIDIA NeMo ecosystem that generate synthetic datasets across text, code, structured, and multimodal data, feeding the pre- and post-training of LLMs such as Nemotron. This role combines hands-on software engineering with applied research in generative methods, and you will collaborate with research, engineering, product, and model teams as well as external labs.

What you'll be doing:

• Build and scale data generation pipelines using LLM-based methods combined with automated quality evaluation. resulting in datasets to improve both initial training and fine-tuning of LLMs such as Nemotron. These data pipelines cover reasoning, coding, structured output, and multimodal understanding.

• Pioneer data generation for agentic and tool-use training: synthetic trajectories, multi-turn interactions, function calling, and executable environments for reinforcement learning, including reward modeling and verifiable-reward data

• Advance multimodal synthetic data generation — image, document, video, and audio — in partnership with NVIDIA's model teams.

• Advance privacy-preserving and safe synthesis — differential privacy, anonymization, and de-identification — enabling model training on sensitive data in regulated domains.

• Develop and maintain open-source libraries and SDKs with clean APIs and strong documentation.

• Drive software excellence with modern tooling, architecture based on configuration, and professional Git/CI-CD.

• Publish original research at top machine learning and AI conferences to maintain NVIDIA's technical leadership.

• Mentor scientists and engineers across the team, raising the technical bar and growing the next generation of researchers.

What we need to see:

• PhD in Computer Science, Machine Learning, Statistics, or a related field, or equivalent experience.

• 15+ years of engineering and research experience in synthetic data generation, generative modeling, multimodal machine learning, or related areas.

• Deep technical understanding of LLMs, how data shapes their pre-training, post-training, and RL stages, and inference frameworks such as vLLM or TGI.

• Proven track record of developing or maintaining software libraries used by a broad developer community.

• Experience building and optimizing scalable data pipelines for large-scale model training — throughput, distributed inference, and cost at cluster scale.

• Strong publication record at premier venues such as NeurIPS, ICML, ICLR, ACL or similar.

Ways to stand out from the crowd:

• Significant open-source contributions in ML or data tooling, with community adoption.

• Experience with multimodal generation or understanding (vision-language, document AI, video, or audio).

• Experience generating data for agentic, tool-use, or reinforcement-learning post-training, including RL environment design.

• Background in differential privacy, de-identification, or synthetic data for regulated industries such as healthcare, finance, or government.

• Experience influencing model training decisions at frontier scale, or partnering directly with pre-training and post-training teams.

NVIDIA is widely considered to be one of the technology world's most desirable employers. We have some of the most forward-thinking and talented people in the world working with us. If you are creative, autonomous, and passionate about building open-source tools that make AI safer and more private, we want t...