Profile cover
H

AI Semantic Scene Graph Construction Engineer

Hannah Cohen

Full-time · Job · San_Francisco

About the role

Understanding a scene isn't just identifying what objects are present — it's understanding the spatial and semantic relationships between them. A scene graph captures that structured understanding: nodes are objects with their properties, edges are the relationships between them (a cup is on the table, the person is holding the cup, the table is in front of the window). Scene graphs enable compositional reasoning and spatial query answering that flat object detection cannot support. We build 3D computer vision systems for robotics and autonomous systems clients. Scene graph construction — building structured relational scene representations from 3D point clouds and RGB-D imagery — is the representation layer that connects our low-level 3D perception (segmentation, reconstruction) to the high-level reasoning that task planning requires. A robot needs to know not just that there is a cup but that it is reachable, upright, and on a stable surface. You'll lead scene graph construction engineering: building the relationship detection models that infer spatial and semantic relations from 3D scene representations, designing the graph schema for robotic task planning use cases, and implementing the real-time scene graph maintenance that updates the representation as the scene changes during task execution.

Responsibilities

  • Design 3D spatial relationship detection models inferring geometric and semantic relations from point cloud representations
  • Build scene graph schema optimized for robotic task planning queries and manipulation reasoning
  • Implement real-time scene graph maintenance updating relations as objects move during task execution
  • Develop compositional scene query answering: 'find all reachable containers that are open and empty'
  • Create scene graph grounding connecting natural language spatial references to graph node identifiers
  • Benchmark scene graph accuracy against human-annotated ground truth for manipulation-relevant relations

Requirements

  • 6+ years in 3D computer vision or robotics perception with scene understanding or graph representation experience
  • Deep understanding of 3D spatial reasoning, point cloud processing, and object relationship modeling
  • Strong Python and PyTorch skills with 3D graph neural networks or relational perception models
  • Experience with RGB-D sensors, 3D reconstruction pipelines, and real-time perception constraints
  • Track record building perception systems that close the loop with planning or manipulation components

Benefits

  • Fully remote with flexible schedule
  • Top-of-market compensation for 3D vision expertise
  • Comprehensive health benefits
  • $5,000 annual professional development budget
  • Equity participation

Keep exploring related roles

Use these topic pages to discover more roles in the same hiring cluster.

Job Type

Full-time

Level

Senior

Language

English

Salary Range

$160k – $200k / year

AI Expertise

AI & Machine Learning Engineers data science & analytics

Skills

python pytorch computer vision hugging face docker gcp

Ready to apply for this role?

Create a free talent account in under 2 minutes.

  • Apply to verified AI companies
  • Get AI-matched job recommendations
  • Message hiring managers directly
  • Build your public AI talent profile

Interested in this role? Create your free profile to apply.

Apply — create free profile