Senior 3D Reconstruction Engineer, Spatial AI

Procore Technologies, Inc.; San Francisco (CA); On-site; USD 141,000 - 194,000; Full time

Job summary

Procore Technologies, Inc. is seeking a Senior/Staff engineer to design and build core 3D reconstruction components and push real-time reality capture for construction-site data.

You will own model architectures, training pipelines, and evaluation benchmarks, turning cutting-edge research into robust, scalable features. You will work hands-on with PyTorch/CUDA, lead technical direction, and collaborate with ML leadership to map a pragmatic roadmap while staying at the frontier of 3D ML and

Qualifications

• Deep, demonstrated expertise in 3D reconstruction and spatial AI.
• Experience turning research into production features at scale.
• Strong grounding in multi-view geometry, SfM, and MVS.

Responsibilities

• Build core reconstruction building blocks: pose, geometry, dense reconstruction, scene representation and rendering.
• Lead evaluation and integration of 3D methods (splitting, neural rendering, foundation models).
• Develop real-time, streaming reconstruction across drone, ground, and 360° data.
• Own architecture, training pipelines, and dataset curation; fine-tune foundation models.

Skills

Gaussian splatting

NeRF-family methods

Differentiable rendering

Feed-forward 3D foundation models

Learned SfM / pose & depth regression

Streaming real-time reconstruction

PyTorch

C++ or CUDA

Education

Master’s or PhD in Computer Science / Electrical Engineering / Robotics

Tools

PyTorch

CUDA

C++

Job description

Procore Technologies, Inc. is seeking a Senior/Staff engineer to design and build core 3D reconstruction components and push real-time reality capture for construction-site data.

You will own model architectures, training pipelines, and evaluation benchmarks, turning cutting-edge research into robust, scalable features. You will work hands-on with PyTorch/CUDA, lead technical direction, and collaborate with ML leadership to map a pragmatic roadmap while staying at the frontier of 3D ML and