Senior Data Quality Engineer, Data Quality Services

Caterpillar Inc.•Irving, TX

1d•$112,710 - $183,140•Onsite

About The Position

Caterpillar is seeking a Senior Data Quality Engineer to define, measure, and automate the quality of training and validation data used for robotics, autonomy, computer vision, and Vision-Language-Action (VLA) systems. Working with a scrum team, this hands-on technical role combines data engineering, machine learning, and quality engineering to develop quality gates, metrics, and validation tools that ensure datasets meet the accuracy, consistency, and coverage standards needed to train and evaluate Caterpillar autonomy models. The engineer will work closely with AI Automation, Data Engineering, perception and machine learning teams, and external annotation providers. The role uses technologies such as NVIDIA Cosmos Curator and Cosmos Reason, along with Caterpillar-specific AI models, to scale AI-driven quality assurance. The focus is on making data quality measurable, automated, and continuously improving at production scale. The position involves multimodal sensor data—including camera and LiDAR data, labels, and captions—from heavy equipment operating on dynamic jobsites. The engineer will define what good data means for Caterpillar's autonomy and Physical AI programs, helping ensure datasets are accurate, consistent, and representative of real-world operating conditions.

Requirements

Bachelor's degree in Computer Science, Data Science, Artificial Intelligence, Machine Learning, Robotics, or a related field.
At least 3 years of experience developing production-grade software using Python and SQL, modern software engineering practices, and containerization platforms such as Docker or Kubernetes.
At least 3 years of experience designing and implementing data pipelines, APIs, and distributed data processing systems.
At least 3 years of experience building data validation frameworks and automated quality checks for large-scale datasets.
Strong understanding of statistics and sampling methods used to audit and measure data quality.
At least 3 years of experience managing camera and LiDAR sensor datasets for autonomy or computer vision, including ground truth generation and annotation quality assurance.
Demonstrated ability to define annotation guidelines, identify labeling defects, and drive measurable improvements in dataset accuracy.
At least 3 years of experience using annotation and dataset management tools such as Labelbox, CVAT, or FiftyOne, and machine learning frameworks such as PyTorch or MLflow.
Strong problem-solving and communication skills, with the ability to work across technical teams and external vendors.
Ability to work full-time at the Irving, Texas office.

Nice To Haves

Experience with computer vision, robotics, autonomy, Physical AI, or multimodal AI systems.
Experience with AI-assisted quality techniques such as label-noise detection, active learning, or consensus-based review.
Experience evaluating vision-language model outputs or working with Vision-Language-Action datasets.
Experience with NVIDIA Cosmos Curator, Cosmos Reason, or similar AI data curation platforms.
Experience with data lineage, dataset versioning, and workflow orchestration tools such as DVC or Airflow.
Experience with cloud platforms, MLOps, and machine learning deployment pipelines.
Experience managing vendor quality programs, including scorecards and SLAs.

Responsibilities

Define data quality standards, acceptance criteria, and annotation guidelines for sensor data, labels, and captions used in perception and VLA datasets.
Build automated quality gates to validate data from ingestion and labeling through dataset publication.
Develop and deploy AI-driven quality checks, including label-noise detection, consensus scoring, and caption verification using vision-language models.
Define and track dataset quality KPIs, including label accuracy, inter-annotator agreement, and rework rate, and share results through dashboards and reports.
Audit internal and vendor annotations using gold datasets and statistical sampling.
Manage annotation providers through scorecards, feedback loops, and SLAs.
Analyze dataset coverage and gaps, trace model failures back to data issues, and ensure dataset lineage, versioning, and governance.