Read cofounder Muhammad Umar's memo, Compression and Reinflation
CodeNinja AtomsUnited StatesSaudi ArabiaPakistan
Unlock free accessTalk to us

CodeNinja Atoms

Sovereign AI for the physical world.

Scroll to explore

CodeNinja Atoms is building a sovereign AI operating system for the physical world.

Sovereign system designs for ports, grids, mills, plants and sites: what to sense, where each model runs, what the object model holds, what it costs and who approves every action.

Designed on Praxis. Made living with Hyper Ontology. Open papers, open object models, open data.

Read the researchPraxisHyper Ontology
44reference architectures
12sectors of physical operations
4countries: Germany, Pakistan, Saudi Arabia, United States
643ontology objects published
44DOIs, all CC BY 4.0

Benchmark · PADI

Physical AI Design Index

Does Praxis produce a better system design than frontier AI on its own? Scored on fresh tasks across 13 industries in three countries.

Praxis platformFrontier AI
0%25%50%75%100%Goal: every checkRSI Loop 12RSI Loop 13RSI Loop 14RSI Loop 15RSI Loop 16RSI Loop 17RSI Loop 18RSI Loop 19RSI Loop 20RSI Loop 2138.4%32.1%39.6%35.7%26.0%16.2%31.3%27.0%30.4%37.7%37.5%29.5%43.2%41.7%36.5%23.9%37.4%41.7%39.1%54.4%RSI Loop 22, blind28.4%48.5%

Loop 22 is a blind evaluation on 7 tasks the platform had never seen. Loops 17 to 21 reused some evaluation tasks, so they are shown for the trend and are not like for like with loop 22.

48.5%vs 28.4%

RSI Loop 22: blind evaluation on seven unseen tasks, Praxis platform vs frontier AI, checks passed

7of 7

Tasks where Praxis came out ahead in the blind evaluation

13of 13

Industries measured across three countries

PADI is the benchmark CodeNinja uses to improve Praxis, one RSI loop (recursive self improvement loop) at a time. Scores are far from saturated: the goal is every check on every task in all 13 industries.

Explore the index

Where Praxis leads · checks passed, all RSI loops

Praxis platformFrontier AI
  • Agriculture and Earth ObservationRSI Loops 14, 20 · 39 checks
    Praxis46.2%
    Frontier35.9%
  • Defense and IntelligenceRSI Loops 13, 15, 20 · 57 checks
    Praxis42.1%
    Frontier33.3%
  • Discrete Manufacturing and AutomotiveRSI Loops 13, 15, 18 · 76 checks
    Praxis43.4%
    Frontier34.2%
  • Energy and UtilitiesRSI Loops 13, 17, 19, 21 · 153 checks
    Praxis36.6%
    Frontier29.4%
  • Maritime and PortsRSI Loops 12, 16, 19, 20 · 114 checks
    Praxis34.2%
    Frontier26.3%
  • MiningRSI Loops 14, 15, 16, 19, 20 · 129 checks
    Praxis42.6%
    Frontier35.7%
  • Oil and GasRSI Loops 14, 18, 19 · 78 checks
    Praxis30.8%
    Frontier26.9%
  • Rail TransportRSI Loops 14, 15, 16, 19, 21 · 134 checks
    Praxis45.5%
    Frontier36.6%
  • SemiconductorsRSI Loops 18, 20, 21 · 57 checks
    Praxis43.9%
    Frontier26.3%
  • Supply Chain and LogisticsRSI Loops 15, 16, 18 · 57 checks
    Praxis31.6%
    Frontier28.1%
  • Warehousing and IntralogisticsRSI Loops 17, 20, 21 · 96 checks
    Praxis38.5%
    Frontier29.2%

All 13 industries

Praxis

Autonomy in physical operations.

CodeNinja Praxis brings that expertise to the hardest problems in physical AI, from mines to test tracks.

Praxis

Praxis in three minutes

Watch it work

Muhammad Umar, cofounder of CodeNinja, walks through Praxis on an open pit mine requirement: the ask goes in, and out come the system design, the architecture, a live simulation and the proposal.

Hyper Ontology

From a reference architecture to a living system.

Every design ships its object model as a package. Hyper Ontology imports it and stands it up over the operator's own systems of record: objects, typed links and actions that sense, decide, act and learn.

How it becomes living

Research · Vertical-Driven Architectures

One operation, one design, end to end.

Each paper is a complete reference architecture for one real operation, written so an engineer, or their coding agent, can build it. Operators are described by class, never by name.

Questions answered by sector: agriculture and earth observation · aviation and aerospace manufacturing · discrete manufacturing and automotive · energy and utilities · heavy industry and construction · maritime and ports · mining · oil and gas · rail and transportation · semiconductors · supply chain and logistics · warehousing and intralogistics

All of them in one table: sovereign AI reference architectures · every object model: industrial ontology examples · by sector: industrial AI reference architecture · the six decisions: physical AI system design · the loop behind them: self-improving AI system · the data model under them: industrial data ontology

Pakistan · agriculture and earth observation

A Self-Improving AI System for Canal Irrigation Water Allocation in Pakistan

Canal Share Ledger: Ontologies for Canal Water Allocation Across the Rotation. A sovereign reference architecture for a provincial irrigation department in Pakistan: one governed canal ontology of sixteen objects over eight read only sources, three drafting agents for shortfall, indent and dispute that named engineers confirm, memory in place of training, and open weights self hosted in the department's own circle server room.

PDF · object model · DOI 10.5281/zenodo.24048587
United States · supply chain and logistics

A Self-Improving AI System for Carrier Exception Handling in the United States

Lane Memory: A Sovereign AI Reference Architecture for Self-Improving Carrier Exception Handling in Supply Chain and Logistics. A self-improving AI system for a third party logistics provider in the United States: one governed freight ontology of sixteen objects over seven read only systems and the working threads, three drafting agents that named coordinators confirm, memory that compounds from every confirmed outcome, and GLM 5.3 self hosted on one node in the operator's own facility.

PDF · object model · DOI 10.5281/zenodo.24048585
Saudi Arabia · heavy industry and construction

A Self-Improving AI System for Cement Kiln Control in Saudi Arabia

Heavy Industry & Construction Watch: One Living Picture of the Operation. A sovereign, read only advisory system for one kiln line at a cement company in Saudi Arabia: a governed ontology of sixteen objects over the plant's seven systems of record, three named agents that draft causes, setpoints and handovers for people to confirm, and memory that compounds confirmed outcomes without ever changing weights, with GLM 5.3 self hosted on one node in the country.

PDF · object model · DOI 10.5281/zenodo.24048545
United States · supply chain and logistics

Industrial Data Ontology for a Cold Chain From Plant to Store in the United States

Lot Thread: Industrial AI for Cold Chain Traceability From Plant to Store. A governed ontology that joins lot, pallet, shipment, receipt and complaint records across five custodians for a refrigerated and frozen food manufacturer in the United States, producing a lot's full journey within hours and excursions joined to exposed lots within the hour.

PDF · object model · DOI 10.5281/zenodo.24048543
Saudi Arabia · aviation and aerospace manufacturing

Physical AI System Design for Composite Part Inspection in an Aviation MRO in Saudi Arabia

Damage Draft: Sovereign AI Reference Architectures for Composite Repair in Aviation and Aerospace Manufacturing. A physical AI system design for the composite repair shop of an aircraft maintenance, repair and overhaul provider in Saudi Arabia: vision at the inspection bench, one governed shop ontology of sixteen objects over six read only systems of record, three drafting agents that certifying engineers confirm and sign, and GLM 5.3 self hosted on one node inside the shop network.

PDF · object model · DOI 10.5281/zenodo.24048561
Pakistan · energy and utilities

Sovereign AI Architecture for Distribution Loss and Theft Reduction in Pakistan

Energy & Utilities Watch: One Living Picture of the Operation. A sovereign reference architecture for an electricity distribution company in Pakistan: one governed ontology of the network and its customers over seven read only systems of record, a loss agent, a tamper agent and a billing agent whose drafts named engineers and officers approve, and GLM 5.3 self hosted on one node in the operator's own data center.

PDF · object model · DOI 10.5281/zenodo.24048559
United States · aviation and aerospace manufacturing

Industrial AI Reference Architecture for an Engine MRO Shop Floor in the United States

Critical Path Watch: Industrial AI in Engine Maintenance on Hardware the Shop Owns. A system design that gives the planners of a United States engine maintenance, repair and overhaul shop serving airlines and lessors one governed model of shop flow, agents that draft while named people decide, on the shop's own hardware with open weights and nothing about the airlines' engines leaving its control.

PDF · object model · DOI 10.5281/zenodo.24048571
Pakistan · heavy industry and construction

Industrial AI Reference Architecture for Fertilizer Plant Reliability in Pakistan

Cycle Watch: Sovereign AI Reference Architectures for Plant Reliability in Pakistan. A reference architecture for a fertilizer manufacturer in Pakistan whose ammonia and urea complex is cycled by gas supply suspensions: one governed plant ontology of fourteen objects over eight read only systems of record and a hardware data diode, three drafting agents that named superintendents and engineers confirm, and GLM 5.3 self hosted on one node on the operator's own hardware in the country.

PDF · object model · DOI 10.5281/zenodo.24048577
Saudi Arabia · supply chain and logistics

Sovereign AI Architecture for a Freight Corridor and Customs Clearance in Saudi Arabia

Corridor Command: Industrial AI for the Whole Consignment Story. A sovereign reference architecture for a national freight and logistics operator in Saudi Arabia: one governed corridor ontology of fifteen objects over seven read only systems of record, three drafting agents that named planners, controllers and service desk agents approve, and open weights on one node in the operator's own data center.

PDF · object model · DOI 10.5281/zenodo.24048557
Saudi Arabia · warehousing and intralogistics

Physical AI System Design for a Mobile Robot Fulfilment Centre in Saudi Arabia

Robot Field Watch: Physical AI for Warehouse Robot Safety at Every Gate, Pick Front and Charging Bay. A sovereign physical AI system that detects people inside the robot field, fallen totes and blocked lanes on the cameras a robotic fulfilment centre in Saudi Arabia already runs, and drafts the zone slowdowns, clearance tasks and handovers that supervisors approve.

PDF · object model · DOI 10.5281/zenodo.24048579
United States · energy and utilities

A Self-Improving AI System for OT Vulnerability Management at a Utility in the United States

Energy & Utilities Watch: One Living Picture of the Operation. A sovereign reference architecture for the OT patch cycle at an electric utility in the United States: one governed fleet ontology of sixteen objects over seven systems of record, three drafting agents that named engineers confirm, memory that compounds every confirmed call, and gpt-oss-120b on one node the utility owns.

PDF · object model · DOI 10.5281/zenodo.24048567
Saudi Arabia · rail and transportation

Sovereign AI Reference Architectures for Passenger Rail Fleet Maintenance in Saudi Arabia

Fleet Condition Command: Industrial AI in Rail Maintenance the Operator Owns. A system design for a national passenger and freight railway in Saudi Arabia that joins two builders' vocabularies into one governed model of the fleet, drafts depot and diagram decisions for named engineers to approve, and computes availability cleanly, on the operator's own hardware and weights inside its own boundary.

PDF · object model · DOI 10.5281/zenodo.24048549
Pakistan · mining

Sovereign AI Reference Architectures for a Remote Mine on a Satellite Link in Pakistan

Mining Watch: One Living Picture of the Operation. A sovereign reference architecture for a remote copper and gold mine on a satellite link in Pakistan: one governed site ontology of fifteen objects over read mirrors of seven systems of record, three drafting agents that named supervisors approve, memory that survives crew rotation, and GLM 5.3 self hosted on one node inside the fence that answers with the link down.

PDF · object model · DOI 10.5281/zenodo.24048581
United States · mining

Industrial AI Reference Architecture for Tailings Storage Facility Monitoring in the United States

Section Story: Sovereign AI Reference Architectures for Mining at One High-Consequence Facility. A sovereign reference architecture for a mining company's high consequence tailings storage facility in the United States: one governed facility ontology of fifteen objects over eight read only systems of record, three drafting agents whose drafts the tailings engineer and the engineer of record confirm, and gpt-oss-120b self hosted on one node of eight 80 GB GPUs in the site server room.

PDF · object model · DOI 10.5281/zenodo.24048569
Pakistan · discrete manufacturing and automotive

Physical AI System Design for Weld and Assembly Quality in a Vehicle Plant in Pakistan

Discrete Manufacturing & Automotive Watch: One Living Picture of the Operation. A sovereign reference architecture for a vehicle assembly plant in Pakistan: triggered cameras and edge units at the station exits checking every body, one governed plant ontology of sixteen objects over six read only sources, three agents that named inspectors and managers confirm, and GLM 5.3 self hosted on one node in the plant's own server room.

PDF · object model · DOI 10.5281/zenodo.24048589
Saudi Arabia · energy and utilities

Sovereign AI Architecture for a Grid Control Room in Saudi Arabia

Energy & Utilities Watch: One Living Picture of the Operation. A sovereign reference architecture for a national grid operator's control room in Saudi Arabia: one governed grid ontology of sixteen objects over ten read only sources, four drafting agents that named engineers and supervisors confirm, and GLM 5.3 self hosted on one node in the operator's own data center.

PDF · object model · DOI 10.5281/zenodo.23889768

Methods

How the designs are made, and how they come alive.

Solutions

Published requirements, solved in the open.

Operators publish what they need. We publish how to build it: the full solution design for a published requirement, from sensing to the living ontology.

  • The requirement as published. What the operator asked for, read in its own words.
  • The complete solution design. Scope, architecture, object model, models and hardware, rollout and cost, reasoned on Praxis.
  • The living system. The object model packaged for Hyper Ontology, ready to stand up over the operator's own systems.

Questions

What this is.

What is CodeNinja Atoms?

CodeNinja Atoms is building a sovereign AI operating system for the physical world. It publishes open reference architectures for physical AI in operations such as energy grids, ports, oil and gas, manufacturing and agriculture, designed on Praxis and made living with Hyper Ontology. CodeNinja Atoms is a fully owned subsidiary of CodeNinja.

What is a reference architecture for physical AI?

A complete system design for one real operation: what to sense, which models run where and on which hardware, the object model (ontology) that joins the operator's existing systems, what it costs over three years, and which named person approves every action.

What does sovereign AI mean here?

The system runs on hardware the operator owns, inside its own country, on open-weight models whose licences let the operator keep and change them, so no operational data has to leave and no single outside AI company sits in the serving path.

What is Praxis?

Praxis is CodeNinja's platform for designing physical AI systems. It turns an operator's requirement into a complete design for physical AI. Every design on this site was made on Praxis. Engineers unlock free access at https://app.codeatoms.ai/signup, and every account is approved by CodeNinja; organisations talk to the commercial team about sovereign and custom deployments.

What is Hyper Ontology?

Hyper Ontology is CodeNinja's ontology platform. It imports the object model a Praxis design publishes (format hyper-ontology/1) and stands it up as a living system over the operator's own systems of record. It is in beta.

What is PADI?

PADI, the Physical AI Design Index, is CodeNinja's benchmark for system design in physical AI. It asks whether Praxis produces a better system design than frontier AI on its own, given the same short requirement, on fresh tasks across 13 industries in the United States, Saudi Arabia and Pakistan. Results are published by industry at https://codeatoms.ai/padi/.

Can I reuse the designs?

Yes. Every paper, object model, model register and dataset row is published under CC BY 4.0 with a DOI. The full set loads as one dataset on Hugging Face, and an MCP server gives coding agents every design.

For agents and engineers

Everything here is data.

Load every design, object, model choice and cost line as one dataset, or read the site the way an agent does.

Praxis

Design for the physical world.

Every design here was made on Praxis. Engineers start free; organisations talk to our commercial team about sovereign and custom deployments.

Free developer access, private by default, your name on the document. Every account is approved by CodeNinja, usually within a day.

Unlock free accessTalk to usSign in

Having trouble? Contact us.