Flag job

Report

Internship 2026 Machine Learning: World Models

Min Experience

0 years

Location

San Francisco, CA

JobType

internship

About the job

Info This job is sourced from a job board

About the role

Bedrock is seeking a motivated intern to help design and train a world model that powers simulation, evaluation, and training for autonomous excavation. Over the course of a 10-12 week internship, you'll prototype architectures, mine real-world data, and develop generative representations to accelerate our autonomy stack. You'll work closely with our Machine Learning team to build models that capture the underlying dynamics of complex environments. Expect hands-on mentorship, regular code reviews, and the opportunity to contribute directly to a foundational piece of Bedrock's ML infrastructure.

About the company

At Bedrock, we've assembled one of the most experienced autonomous technology teams in the industry, with deep expertise scaling breakthroughs across transportation, infrastructure, and enterprise software. Our leaders helped put the first self-driving cars on public roads at Waymo, scaled systems for Segment's $3.2B acquisition, and grew Uber Freight to $5B in revenue. While others debate the future of AI, we're deploying it in the real world. Our systems are already installed on heavy machines across the country, learning on real construction sites and working to reshape the earth with survey-grade precision and exceptional safety. This isn't a simulation—it's autonomous intelligence working on billion-dollar infrastructure projects. In just over a year, we've raised $80M, put our equipment into the field, and established partnerships with forward-thinking contractors who are integrating our technology into their operations. We're working quickly to close the gap between America's surging demand for housing, data centers, manufacturing hubs, and the construction industry's growing labor shortage.

Skills

machine learning
simulation
autonomy
world models
python
deep learning
pytorch
jax
tensorflow
reinforcement learning