Staff ML Engineer - Embodied AI Scaling Foundations

GM • Sunnyvale, CA

18h

About The Position

As a Staff AI/ML Engineer in General Motors’ Embodied AI Data Foundations organization, you will work as an individual contributor developing data-centric AI solutions to improve autonomous driving performance. You will design and run data curation and model-training recipes that use real and synthetic data to produce models capable of safe, reliable behavior across diverse real-world scenarios. The Data Scaling team owns the data flywheel for autonomous-vehicle foundation-model pre-training and supervised fine-tuning, defining the data and quality needed to improve driving behavior at scale. The team works with large existing datasets and helps shape future high-value data collection, aiming to increase the effective data behind each major model release.

The role focuses on understanding how data composition affects model behavior, developing curation and mining methods, and defining evaluations that predict on-road performance. You will train models at scale, investigate failures and their data causes, collaborate with platform and product teams to bring models into onboard driving systems, and share findings and best practices. The team also studies data scaling for embodied driving and supports publishing relevant findings.

Requirements

• Master's or PhD in Computer Science, Robotics, or Machine Learning.
• Strong machine-learning fundamentals, including the ability to design experiments, select baselines, interpret ablations, and distinguish signal from noise.
• Proficiency in Python and PyTorch, with experience training models on large datasets.
• Hands-on experience with data-centric machine learning, such as curation, sampling, labeling, or evaluation of large training sets.
• Working knowledge of large-scale foundation models and their pre-training, fine-tuning, and alignment.
• Data analysis skills with NumPy and Pandas; SQL or Spark for large datasets.
• Demonstrated ability to deliver applied machine-learning results under real-world constraints and timelines.
• Clear communication skills, including explaining results and their limitations to engineers and non-experts.

Nice To Haves

• PhD, publications, or open-source contributions in representation learning, multimodal or vision-language models, generative models, reinforcement learning, or data-centric machine learning.
• Experience with robotics, autonomous driving, or other embodied AI systems.
• Experience with synthetic and simulation data, including sim-to-real transfer.
• Familiarity with production machine-learning deployment workflows.

Responsibilities

• Design and run experiments linking data composition to model behavior, including dataset mixtures, sampling strategies, curricula, and scaling-law studies.
• Apply self-supervised pre-training, imitation learning, reinforcement learning, and foundation-model fine-tuning to driving behavior, trajectory generation, and perception tasks.
• Develop data curation and mining methods, including auto-labeling, deduplication, difficulty and uncertainty estimation, and discovery of long-tail and out-of-distribution scenarios.
• Define offline metrics and evaluations that predict on-road behavior, and use evidence from them to guide model and data decisions.
• Trace model failures to their root causes in data and specify the data needed to address them.
• Train models at scale across large multi-GPU and multi-node datasets, partnering with platform teams on required pipelines and tooling.
• Collaborate with cross-functional teams to bring models into onboard driving systems.
• Document learnings and best practices.
• Follow relevant literature and incorporate promising advances into training recipes and evaluations.

What This Job Offers

• Job Type: Full-time
• Career Level: Senior