Pseudo-Labeling and Data Pipeline Engineer

What You'll Do:

Design, implement, test and deploy tooling and pipelines for internal quality control and issue identification of pseudo-labeled data, ensuring annotations meet the bar required by downstream model training.

• Provide statistical support and report on pseudo-label quality, coverage, and pipeline health to internal stakeholders and leadership.
• Support secondary data selection (virtual packaging) to curate and feed on-demand data to online model training.
• Support the build-out of the ML data delivery system that enables the online perception team to train models on-demand.
• Drive general pipeline improvement and optimization across the pseudo-labeling and data delivery stack, identifying and resolving bottlenecks in throughput, quality, or reliability.
• Demonstrate project management skills, serving as project lead guiding less experienced team members in multiple facets of project execution.
• Stay up to date with the latest developments in offline perception, data pipeline engineering, and ML data infrastructure for autonomous driving.
• Independently develop tools, services, and algorithms using disciplined software development processes, making recommendations for developing new code or re-using existing code, implementing version control, and maintaining documentation of created applications.
• Define and implement ingestion, data preparation, curation, and governance of large, multi-faceted data sets supporting analytics and ML training workflows.
• Proactively assess current capabilities to identify areas for improvement, proposing solutions that align with core strategy and operation.
• Guide and produce information products, supporting visualization and data accessibility in a customer-centric manner.
• Evaluate and make recommendations regarding technical advances that improve productivity and quality, reduce flow times, and enhance operational surety.
• Develop guidelines and standards for data quality control, data delivery systems, and their deployment, and associated processes.
• Provide technical guidance or business process expertise, technical leadership, coaching and mentoring to team members.

What You'll Need to Succeed:

Considered highly skilled and proficient in discipline; conducts complex, important work under minimal supervision and with wide latitude for independent judgment.

• Scope of Influence: Expected to drive alignment across team interfaces to the rest of the organization. Designs, maintains and owns team technical solutions and drives consensus. Mentors and guides engineers within the group.
• Bachelor’s Degree in Computer Science, Robotics, Electrical Engineering or related technical field plus demonstrated competencies and technical proficiencies typically acquired through 6+ years of experience OR;
• Master’s Degree in Computer Science, Robotics, Electrical Engineering or related technical field plus demonstrated competencies and technical proficiencies typically acquired through 3+ years of experience OR;
• Required Qualifications (some combination of the following skills):
  • Familiarity with the offline perception stack in general, and knowledge of how pseudo-label data is produced and related best practices.
  • Strong software engineering background building and operating data pipelines and services at scale.
  • Statistical analysis and reporting skills, with the ability to translate data quality findings into actionable insights.
  • Scaled MLOps and Tooling – ML Frameworks, experiment tracking, model registry, MLflow, Weights & Biases, ML Metrics and Evaluation / Quality.
  • Model Data Curation – Parquet data processing (PyArrow, Daft, Pandas, etc).
  • Development Tools & Eco-System (at scale) – Proficiency in Python software development. Also, VDI and cloud-based development environments, CI Systems (GitHub Actions), and Docker.
  • Experience with distributed data processing and/or ML frameworks – PyTorch, Lightning, Ray, or similar.

Bonus Points!

• Experience with large-scale data delivery systems and associated quality control.
• Pseudo-labeling experience in general.