Senior Applied Research Scientist, Data Curation
$224k - $357k • Remote • US, CA, Santa Clara
Posted 1mo ago
Job Location
US, CA, Santa Clara
Tech Stack
Remote Work Policy
Fully remote
Categories
AI Research Engineer
About the job
NVIDIA is seeking a Senior Applied Research Scientist to join its Curator team. This role focuses on researching, developing, and deploying deep learning models and data pipelines at scale for foundation model training. You will work on next-generation data curation and document extraction, contributing to capabilities that power NVIDIA Nemotron Parse, with an emphasis on structured content extraction, data quality, and deduplication for petabyte-scale datasets. The team is building the foundations for modern LLMs, playing a critical role in training leading open LLM models.
Responsibilities
- Develop efficient models and data pipelines for extracting and curating multimodal data (documents, images, audio, video) for foundation model training.
- Build petabyte-scale extraction and content deduplication pipelines, including document/HTML parsing and various deduplication methods (fuzzy, near-duplicate, semantic, substring).
- Expand and optimize curation methodologies for petabyte-scale multimodal data on large GPU clusters to enhance foundation model training sets.
- Develop standard research methodologies by crafting datasets, metrics, experiments, and validation scripts.
- Assist ML Engineers in scaling pipelines to production by developing NVIDIA Inference Microservices (NIMs) and deployment blueprints.
- Publish research findings through papers, blog posts, documentation, and training materials.
- Stay updated on the latest advancements in data curation research within academia and industry.
Requirements
- Master's, Ph.D., or equivalent experience in data curation, document AI, information retrieval, or multimodal research.
- Track record of publications in leading conferences (e.g., CVPR, ICCV, ECCV, KDD).
- Hands-on experience developing computer vision and document-extraction models/pipelines (layout analysis, OCR, table/figure/formula extraction).
- Strong understanding of state-of-the-art data curation research, focusing on multimodal content extraction and deduplication.
- 10+ years of experience developing multimodal systems across various models and platforms.
- Proven expertise managing distributed data frameworks (e.g., Ray, Spark, Dask).
- Experience deploying massive, multi-node ML or data processing tasks in production.
- Knowledge of best practices for batching, streaming, and scaling ingestion pipelines.
- Excellent Python programming skills.
- Strong hands-on experience with PyTorch or comparable deep learning frameworks.
- Ability to clearly communicate ideas through publications, code repositories, etc.
- Excellent communication and interpersonal skills, with the ability to work in a dynamic, distributed team.
- Kaggle Grandmaster status or strong record in ML competitions is a strong plus.
- Information retrieval experience is a big plus.
- History of mentoring junior engineers and interns is a plus.
Benefits
- Highly competitive salaries
- Comprehensive benefits package
- Equity