FieldAI
Field AI · Inspection · Shipping since 2024
Last updated: 2026-04
Summary
FieldAI is a legged-robot autonomy platform that pairs a sensor-compute payload with a Boston Dynamics Spot to run autonomous site inspections on active construction projects. It covers six documented subtasks across safety monitoring, daily walkthroughs, 3D scanning, BIM-to-field comparison, and defect tracking; the superintendent and safety team still own interpretation, response, and corrective action. FieldAI has been shipping since 2024 with four confirmed construction deployments across two named customers (DPR Construction and Big-D Construction) and two unnamed top-10 ENR firms.
How it works on your site
PPE Compliance Monitoring
The robot traverses active floors and identifies workers missing required PPE, then flags violations to the project team. The safety manager still reviews each alert and determines the response.
Hazard Identification and Unsafe Behavior Detection
The system uses onboard vision and LiDAR to scan for physical hazards: standing water, blocked egress routes, holes, drop-offs, slippery surfaces, and machinery proximity. The safety team reviews flagged conditions and decides on remediation.
Daily Walkthrough and Progress Documentation
The robot runs scheduled or on-demand documentation passes on a daily cadence, capturing photo and scan data across each floor. It replans its path when site conditions change between runs. The superintendent reviews the captured data rather than walking the floors for routine documentation.
3D Scanning and Point Cloud Capture
The robot generates 360-degree scans using LiDAR and vision sensors, processing all data on-edge without cloud connectivity. NVIDIA Omniverse NuRec handles 3D reconstruction from the captured data.
BIM-to-Field Comparison (Scan-vs-BIM)
The system overlays captured scans against BIM models to identify deviations from the design intent. The robot navigates without prior maps or GPS, relying on SLAM and on-edge processing to build its own spatial reference as it moves through the site.
Defect Identification and Punch List / Closeout Tracking
The robot captures daily scan data that feeds defect identification and progress tracking against the project schedule. The project team still owns the punch list workflow, interpretation of flagged items, and closeout documentation.
How it worked on other sites
DPR Construction at Data Center
- Production deployment running approximately 1.5 years as of November 2025.
- Tasks include progress documentation, multi-floor mapping across 4 floors, 125,000 sq ft of roofing surveys, 500,000 sq ft of interior scanning, hazard detection, and security checks.
Two unnamed top-10 ENR construction firms
Two unnamed top-10 ENR construction firms- Deployment running over two years as of February 2026 across Asia, Europe, and North America.
- Tasks include inspections, mapping, monitoring, 3D scans, hazard identification, and overnight operations.
Ryan Companies US, Inc. at ATX Tower (58-story mixed-use)
- Single-event demonstration of autonomous reality capture at a 58-story mixed-use tower in Austin, Texas.
Robot also used by
