AROBOTIX: Technology

AROBOTIX is an AI-native system with engineered guardrails. The wedge is a compiler that turns models into structured, verifiable work. Everything else is built on top of it.

Foundational wedge Compiler before UI

BIM-to-Executable Task Compiler

The first critical system. It reads the digital model and compiles it into structured execution tasks, with readiness validation, sequencing, and dependency mapping built in. It turns drawings into instructions without dictating means and methods.

In the Crawl phase, classification is rule-based, which is sufficient and more defensible for the pilot than a black-box model. The output is clear, verifiable, trackable tasks.

Readiness validation logic and dependency mapping
Sequencing options, not sequencing commands
Revit support via Autodesk Platform Services, open BIM via IFC

# BIM model → structured executable tasks

task = compiler.compile(model="tower_L12.ifc")

{

  "task_id": "STR-V-0421",

  "group": "STRUCTURE-VERTICAL",

  "readiness": ["rebar_set", "formwork_ok"],

  "depends_on": ["FND-0188"],

  "assignable_to": ["human", "machine"],

  "verify": "proof_of_work"

}

Architecture Locked first

Construction Task Object Model (CTOM)

The schema that everything else depends on. CTOM is organized by execution pattern, not trade name, so the same logic generalizes across projects. Each group carries its own readiness prerequisites, verification requirements, and dependency structure.

Getting this schema right is the difference between a system that compounds and one that breaks on the second project.

10 execution-pattern field groups

01Site-Civil

02Foundations

03Structure-Vertical

04Structure-Horizontal

05Envelope

06MEP-Rough

07Interiors-Construction

08Finishes

09MEP-Finish

10Specialty-Systems

Hero output Pilot deliverable

Proof-of-Work Record

A structured, confidence-tagged record that a task was executed: the time, the machines, the conditions, and the human confirmation, all traceable. It is a defensive record and a dispute-reduction layer.

It is not automated QA signoff, legal certification, or fault assignment. Human review and final accountability stay central. When a question comes up months later, the verified record surfaces in seconds, before it becomes a claim.

Confidence-tagged, time-stamped, traceable
Built around contractor authority, not against it
Turns disputes into lookups

Record · concrete pour confirmed

taskSTR-H-0912 · slab pour L12

timestamp2026-08-04 10:03 ICT

assetspump-02, vibrator-04

conditions31°C · slump verified

confirmed_bysite supervisor

confidenceHIGH · 0.94

The moat Compounding

Execution Data Graph

Every project we run produces a dataset no competitor can replicate: the first verified, confidence-tagged record of how construction actually happens at task level. It grows with every build and gets harder to replicate with every cycle.

Features can be copied. Operational intelligence compounds. The graph powers benchmarking, risk prediction, and the pattern recognition that makes the next project's coordination smarter.

Compounding data loop

Better coordination

readiness and dependency resolved per task

↓

Cleaner execution data

confidence-tagged Proof-of-Work records

↓

Smarter coordination

graph traversal: readiness propagation, blocking cascades

↓

Stronger network effects

benchmarking and risk prediction across projects

↻ compounds every project cycle

Neutral infrastructure Robot-agnostic

Orchestration & AI-Assist

A neutral intelligence layer that delivers contextual insight across different robots and applications. It captures real-time execution data, sequences work to site conditions, and answers project questions by citing the underlying project data.

AI assists task decomposition with human validation checkpoints throughout. Every robot vendor owns its silo. No one orchestrates across all of them. That neutral position is the strategic advantage, and taking sides with any OEM would destroy it.

Robot-agnostic APIs and partner integrations, no hardware lock-in
AI-assisted decomposition with human checkpoints
Answers cite project data, not guesses

Stack

apiFastAPI

dataPostgreSQL

cloudAWS

bim_revitAutodesk Platform Services

bim_openIFC schema

classifyUniclass · OmniClass

Hardware on the Roadmap

Software first. Hardware where it compounds.

Planned field devices that extend the platform from the screen to the slab. Each one feeds the same execution data graph, and none is required for the software to work. We lean toward integrating proven hardware rather than building from scratch. These are exploratory roadmap items, not shipping products.

Field UI Viability: High

AR Readiness & Model-Viewing Headset

A heads-up display that overlays the compiler's task sequence, readiness status, and model geometry onto the physical work area. Most likely an integration play on existing AR hardware rather than a ground-up build, which keeps it doable.

It is the field-facing window into the software: the place where the plan meets the slab.

Overlays task sequence and readiness on-site
Integration-first, not a ground-up build
The crew's view into the compiler