Staff support & internal logistics
Deliveries · errands · internal transport · routine runs
Reduce non-care interruptions and give staff more capacity for care.
- Field-tested in Swiss healthcare environments
- On-device architecture
- Built in Zürich
The bottleneck Robots can navigate hospitals.
Robots can navigate hospitals.
Adoption still stalls.
01 Robots still behave like machines
They hesitate, block people, misread social situations and require staff attention.
02 Healthcare workflows are not generic
A technically functional robot is not automatically operationally useful.
03 Staff need automation, not another system to manage
Repetitive support work still pulls teams away from care.
Navigation gets a robot from A to B. Reliable behavior determines whether people actually use it.
“We did not underestimate the robots. We underestimated what it takes to make them work inside a hospital.”
The system
One system, engineered across three layers.
Healthcare providers receive an operational robotic system. VRS controls the sensing, compute and behavioral layers that determine how it works around people.
Field requirements become reusable product improvements — not one-off customer builds.
- 01 Mobile robot platform Replaceable
Industrial mobility base. Replaceable by design.
- 02 VRS sensing & compute VRS-controlled
Perception · sensors · edge compute · interfaces
- 03 CareOS VRS-controlled
Behavior Engine · safety · human-aware navigation · interaction
Healthcare deployment
Workflow configuration · site validation · monitoring · iteration
Applications
Start with repetitive staff-support workflows.
01 Guidance
Reception, wayfinding and escorting visitors to the right ward.
02 Interaction
Spoken and on-screen interaction, escalating to staff when unsure.
03 Patient / resident engagement
Light social and cognitive activity on the ward.
Primary deployment KPI → staff time returned through automation
Real-world deployment
From staff request to completed mission.
- 01 Request
- 02 Mission
- 03 Perception
- 04 Behavior selection
- 05 Navigation / interaction
- 06 Completion
- 07 Replay
Recorded mission On-robot log
A real mission replayed from on-robot logs: floor geometry, the driven trajectory, people as anonymized geometry, and the live behavior mode with its reason code.
Every mission is recorded decision by decision, with a plain-language reason for each one, and can be replayed in 3D.
Field validation
Completed deployments Built in the field, not just in the lab.
Haus Tabea
✓ 
Swiss Paraplegic Centre
✓ Cross-floor mission completion 2/12 → 2/2
Same building, same route, one software change. The controlled before-and-after of a single fix.
Safety record, 47 missions 0 contacts
No person or wheelchair contact. Six near-misses, each corroborated against LiDAR.
Critical failure mechanism 4 days
Identified from mission logs and corrected in the field within four days.
Next deployments
Request the field report → Two signed letters of intent with Swiss care providers. Scheduling and funding are in progress; we name a site only once it is contracted.
Technology
The technology behind the system.
CareOS
Behavior Engine
- Deterministic execution
- Fail-closed safety logic
- Human-aware navigation
- Interaction orchestration
- Mission replay
- Explainable decisions
VRS sensing & compute
Deployed today in service
- Slamware platform abstraction
- ZED 2i stereo depth with LiDAR fusion
- Jetson Orin Nano compute on the robot
- On-robot tablet interface; voice is the one cloud-assisted function
Next platform · robot #2 in design
- Jetson AGX Orin
- Two additional cameras
- Modular payload back
Both renders are design work for robot #2, not the robot in service today.
Learning loop
Every deployment improves the next.
- 01 Deploy
- 02 Observe
- 03 Replay
- 04 Diagnose
- 05 Validate
- 06 Update
01 On-device processing
Perception and behavior run on the robot itself, not in a data centre.
02 Mission-level explainability
Every decision carries a reason code and can be replayed in 3D.
03 Versioned behavior updates
Behavior changes ship as versioned updates, validated against recorded missions before they reach a site.
04 Privacy-conscious architecture
Video and perception are processed on the robot. Voice interaction is the one function that uses the cloud; no video or images leave the building.
Company
Core team A robotics company, built in Zürich.
Salim Benamira Founder & CEO Ex-senior consultant, PwC
Anatole Conrad Co-founder & CTO Software engineer & ML specialist (ex-ENSICAEN)
Seomi Lee Research engineer · representation learning ETH Zürich · multimodal self-supervised scene representation
Prof. Dr. Yulia Sandamirskaya Scientific advisor Professor of robotics & neuromorphic computing, ZHAW
Clinical partners
Academic partners
Programs & infrastructure
Deployment
Explore VRS in your care environment.
We start with your workflows, identify where robotic automation can create measurable value, and define the right evaluation format.
How evaluations work
- 01 Workflow review
A short session on your actual routines, volumes and constraints.
- 02 Site assessment
Floors, elevators, doors and corridor traffic reviewed on site.
- 03 Scoped evaluation
A defined evaluation format with agreed metrics, duration and scope.