Hardware
DPU / secure appliance lead
BlueField, DOCA, SmartNIC, secure boot, and hardware-adjacent enforcement.
Company
A hardware-first AI safety company for autonomous machines.
Celestial State is building the trust anchor for physical AI: the layer that authorizes machine actions, signs proof, and gives OEMs and operators evidence they can defend.
Thesis
The next AI safety problem is physical.
01
Physical AI is becoming infrastructure.
Autonomous machines are moving from demos into warehouses, factories, yards, roads, hospitals, and energy systems.
Autonomous machines are moving from demos into warehouses, factories, yards, roads, hospitals, and energy systems.
02
The proof layer is missing.
After an incident, teams need to prove what the machine was authorized to do, what it did, and which policy governed it.
After an incident, teams need to prove what the machine was authorized to do, what it did, and which policy governed it.
03
Hardware makes it defensible.
The most valuable proof cannot live only in the software stack controlled by the machine it is meant to govern.
The most valuable proof cannot live only in the software stack controlled by the machine it is meant to govern.
Founder
Moshe
Khorshidi
CEO & Technical Founder · Tel Aviv, Israel
Senior data Platform Engineer and technical lead with production experience in cloud infrastructure, data pipelines, and distributed systems.
Celestial State starts with the software control plane and expands into hardware depth as the pilot validates the core product. The thesis is that high-consequence machine actions need a trusted authority point the machine itself cannot rewrite — and that proof must be hardware-rooted to be defensible.
contact@celestial-state.com
Proof of work
01Physical AI governance thesis
02Reference architecture
03CAGP decision protocol
04Demo storyboard
05AWS startup program acceptance
06Israel Innovation Authority POC application
Hiring Plan
The company becomes hardware-native by design, not by slogan.
Security
Security architect
Signing, attestation, protected keys, tamper-resistant logs, and trust boundaries.
Robotics
Robotics integrations
ROS2, PLC, OPC UA, fleet managers, industrial networks, and machine controllers.
GTM
Design partner lead
Pilot discovery with OEMs, operators, system integrators, insurers, and regulated customers.
Position
What we are building.
A secure hardware trust layer for physical AI — starting with signed action receipts and moving toward real-time enforcement.
This is not a monitoring dashboard or a model safety wrapper. The core claim is straightforward: machines that can cause physical harm need a trusted, independent authority that can approve, block, and record every high-stakes action.
Tel Aviv, Israel
Company building