Simulation Intern: Neural Rendering for Real-World Autonomy

Bedrock Robotics Inc; San Francisco (CA); On-site; USD 34,000 - 62,000; Full time

Job summary

Bedrock Robotics is seeking a Simulation Intern to help push neural rendering for our autonomous construction-site stack. You’ll develop and benchmark rendering methods, and integrate outputs into the sim pipeline to support perception and planning teams.

You’ll collaborate with a world-class team, apply Python and PyTorch on real-world site data, and tackle challenges like dust, lighting, and dynamic objects to close the visual gap that affects autonomy performance.

Qualifications

• Currently pursuing a BS, MS, or PhD in computer science, computer graphics, robotics, or a related field — or bringing equivalent research or industry experience.
• Strong Python and hands-on experience with PyTorch (or equivalent)
• Solid understanding of 3D graphics fundamentals: rendering pipelines, scene representations, camera models, and coordinate systems
• Familiarity with neural scene representations (NeRF, Gaussian splatting, neural radiance fields, or related methods)
• Comfort with real-world data — construction sites are messy, dusty, and nothing like an indoor dataset

Responsibilities

• Research, implement, and benchmark neural rendering methods (NeRF, 3D Gaussian splatting, or related techniques) for generating realistic construction-site imagery within our simulation stack
• Train models on real-world site data captured by our fleet and evaluate visual fidelity, temporal consistency, and render speed
• Integrate neural rendering outputs into the simulation pipeline so perception and planning teams can train and test against them
• Build evaluation metrics and tooling that quantify how well rendered scenes match real fleet data — and whether that improvement transfers to autonomy performance
• Identify and address failure modes: dynamic objects, deformable terrain, dust, harsh lighting, and other construction-specific challenges
• Document findings, limitations, and a recommendation on where neural rendering should (and shouldn't) replace traditional sim assets

Skills

Python

PyTorch

3D graphics

Neural rendering

Benchmarking

Education

BS/MS/PhD in CS/CG/Robotics

Tools

Unreal

Unity

NVIDIA Isaac Sim

CUDA

Job description

Bedrock Robotics is seeking a Simulation Intern to help push neural rendering for our autonomous construction-site stack. You’ll develop and benchmark rendering methods, and integrate outputs into the sim pipeline to support perception and planning teams.

You’ll collaborate with a world-class team, apply Python and PyTorch on real-world site data, and tackle challenges like dust, lighting, and dynamic objects to close the visual gap that affects autonomy performance.