PhD Research Intern: Synthetic Data for Autonomous ML

Waymo LLC; Mountain View (CA); On-site; USD 99,000 - 135,000; Full time

Benefits offered by this job

Generous benefits programs

Internship opportunity

Job summary

Waymo LLC is offering an internship for PhD students in a cutting-edge AI Foundations role focused on autonomous driving data. You will work within the AI Foundations team to address open problems in reinforcement learning, learning from demonstration, and robust evaluation, applying your research to practical driving data.

The role emphasizes collaboration with industry leaders, hands-on model training with synthetic and logged data, and presenting methods for synthetic data generation and

Qualifications

• Currently pursuing a PhD in Computer Science, Machine Learning, Robotics, Electrical Engineering or a related technical field.
• Familiarity with methods of sampling from large models.
• Experience with post-training methods such as SFT, RL, and preferences.
• Modern deep learning frameworks such as PyTorch and Jax.

Responsibilities

• Build tools for sampling, selecting, and discarding synthetic driving data with testable hypotheses of how that data will lead to improvement
• Train and evaluate models with mixtures of synthetic and logged data
• Present and write about methods for synthetic data generation and training

Skills

Sampling from large models

Post-training methods (SFT, RL, prefs)

Open and closed loop decision making

Education

PhD in CS/ML/Robotics/Electrical Engineering or related field

Tools

PyTorch

JAX

Job description

Waymo LLC is offering an internship for PhD students in a cutting-edge AI Foundations role focused on autonomous driving data. You will work within the AI Foundations team to address open problems in reinforcement learning, learning from demonstration, and robust evaluation, applying your research to practical driving data.

The role emphasizes collaboration with industry leaders, hands-on model training with synthetic and logged data, and presenting methods for synthetic data generation and