Hardware
Ultra-compact drones
Sub-250-gram, C0-class compliant aircraft cleared to operate in indoor and outdoor environments without heavy regulatory friction.
EPFL Spinoff · Lausanne, Switzerland
Robotic safety intelligence
for high-risk worksites
InspectNet combines autonomous robotics with regulation-compliant AI to deliver real-time hazard detection and auditable reports across construction sites, power plants and factories.
- >95%detection accuracy
- 50×faster than manual
- 80%cost reduction
- 24/7monitoring
Backed by
EPFL
BRIDGE
Venture Kick
The product
End-to-end automated safety inspection
High-risk industries lose lives, time and capital to preventable safety lapses. InspectNet closes that gap by combining advanced robotic data acquisition with regulation-compliant data processing to generate hazard alerts and auditable site reports in real time.
Built privacy-by-design, with anonymization embedded throughout the pipeline.
-
Real-time hazard alerts
Notifications reach safety teams the moment a risk is detected.
-
Auditable site reports
Generated automatically at configured intervals, ready for regulators.
-
Non-disruptive capture
Robots collect structured data without interrupting operations.
-
Regulation-compliant
Pipelines aligned to industry safety standards out of the box.
How it works
Three integrated stages
From the rotor blades to the report. One continuous pipeline.
-
01
Data acquisition
Autonomous robots capture structured, high-quality data without disrupting operations on site. User input is seamlessly integrated into the pipeline.
-
02
Intelligent processing
An AI pipeline identifies safety risks and anomalies in near real-time across 2D and 3D scenes.
-
03
Alerts & reporting
Immediate notifications to safety teams, and auditable reports generated at configured intervals.
Technology
Robotics and AI, engineered together
Hardware
Indoor & outdoor docking
Compact docking stations require minimal infrastructure, enabling persistent autonomous operation on any site.
Coordination
Swarm-enabled throughput
Multi-robot coordination delivers 5× the throughput of a single unit, covering more area in less time.
AI & computer vision
Detection & segmentation
Object detection, instance segmentation and 3D spatial analysis pinpoint hazards across 2D imagery and reconstructed scenes.
AI & computer vision
Vision-language models
Fine-tuned VLMs reason over scenes the way an inspector would, flagging context-specific risks, not just objects.
Privacy
Privacy by design
Anonymization is built into the pipeline, keeping people on site safe and identifiable data out of the model.
What we detect
Nine categories of site hazards
A single pipeline, trained on the failure modes that matter.
- PPE non-compliance
- Unprotected edges
- Danger-zone violations
- Unstable objects
- Missing signage
- Machinery proximity
- Fire risks
- Temporary structure failures
- Struck-by hazards
Performance
The numbers behind the pipeline
>95%
Detection accuracy across the supported hazard categories.
50×
Faster than manual safety inspection workflows.
80%
Cost reduction relative to traditional inspection approaches.
24/7
Continuous monitoring, with no shifts and no downtime.
Where we are
Incubated at EPFL, Headed for MVP
Raven Intelligent Systems is currently under EPFL incubation, finalizing prototypes and advancing toward a MVP release. We have started our pre-pilot study at EPFL Double Deck construction site and actively looking for pilot partners across construction, energy and heavy manufacturing.
Become a pilot partner- Today
Prototypes in iteration. Pre-pilots started. Pre-seed stage.
- Next
MVP release with full three-stage pipeline.
- Beyond
Swarm-scale deployments across construction, energy & manufacturing with expanded hazard coverage.
Get in touch
Talk to the team.
Whether you operate a high-risk site, lead a safety program or simply want to understand the technology, we're listening.