Flagship Platform · Autonomous Construction Intelligence

AI that prevents construction incidents — not just reports them.

ACIP is a full-stack intelligence platform for construction safety, powered by the Live Risk Intelligence Engine (LRIE). It fuses computer vision, machine learning, document intelligence, and multi-agent orchestration into one deterministic decision engine that observes, scores, decides, and escalates — closing the loop from detection to action without a human bottleneck.

1 in 5

Pioneering safety intelligence at a magnitude the industry hasn't seen.

Construction is the most dangerous major industry — one in five workplace fatalities happens on a site. Today's tools either report incidents after they happen or display yesterday's data in dashboards. ACIP is built to be the first platform that spans every phase of construction, fuses every signal, and acts in seconds — preventing the incident instead of documenting it.

Why ACIP is different

Three structural advantages, one closed loop.

🔄

Closed-loop intelligence

Detection to action with no human in the bottleneck. LRIE detects, scores compounding risk, retrieves the right protocol, decides, and escalates to the right person in seconds — every step logged.

🧬

Multi-modal risk fusion

Not "missing hard hat = HIGH." Instead: missing hard hat + 12th floor + 20 mph wind + this trade's three violations this week + a crane pick in 30 minutes = CRITICAL, with specific mitigation steps.

🧠

Institutional memory

Projects are temporary; their lessons usually leave with the people. ACIP's vector store and trained models turn every project's learnings into persistent organisational memory applied to the next site automatically.

Full lifecycle coverage

Intelligence across every phase of construction.

ACIP's deepest capability is live execution monitoring — but its document and predictive modules extend risk intelligence upstream to planning and downstream to closeout.

01
Planning
02
Design
03
Procurement
04
Pre-Construction
05
Execution
06
Closeout
Phase 01 · Planning

Pre-construction risk baselines

  • Geographic risk factors — flood zones, seismic areas, underground infrastructure
  • Contractor risk profiles from prior safety performance
  • Seasonal risk windows — heat stress, winter hazards
How: predictive risk module (XGBoost) analyses historical incident data and site metadata.
Phase 02 · Design

Design-stage hazard discovery

  • Design elements that correlate with incidents — confined spaces, elevated platforms, crane swing radii
  • Conflicting specs between architectural and structural plans
  • Missing safety provisions in temporary works
How: Document Intelligence (NLP + RAG) parses BIM, structural, and MEP plans against safety protocol databases.
Phase 03 · Procurement

Trade partner risk validation

  • Subcontractors with elevated incident rates or expired certifications
  • Gaps in required safety training for trade scopes
  • Non-compliant insurance for high-risk activities
How: document pipeline extracts and validates safety data from procurement packages.
Phase 04 · Pre-Construction

Subsurface & existing-condition risk

  • Unmarked underground utilities and subsurface infrastructure
  • Environmental hazards — asbestos, lead, contaminated soil
  • Inadequate traffic control for urban sites
How: cross-references as-builts, utility surveys, and geotech reports into a subsurface risk map.
Phase 05 · Execution — Primary focus

Live jobsite monitoring

  • Real-time PPE violations — hard hats, vests, harnesses, glasses, gloves
  • Unsafe practices — unprotected edges, improper scaffolding, exclusion-zone breaches
  • Compounding-factor risk escalation and trend-based alerts
How: full LRIE loop — vision (YOLOv8) → risk scoring (XGBoost) → conditional RAG → escalation.
Phase 06 · Closeout

Portfolio learning capture

  • Recurring safety issues that should inform future planning
  • Incomplete safety documentation creating liability exposure
  • Near-miss patterns indicating systemic issues
How: LRIE audit logs are mined for portfolio insights; lessons vectorised into ChromaDB for future retrieval.
Live Risk Intelligence Engine

A deterministic brain behind every decision.

Every AI output — vision, ML, or document — passes through LRIE's state machine before any action is taken. Each transition is typed, logged, and auditable for OSHA compliance and insurance.

DATA_RECEIVED
Observe

Ingests site camera images, inspection photos, PDF reports, CSV logs, IoT streams, and manual entries. Each input is typed and routed to the right AI module.

PREPROCESSING
Validate

Pydantic v2 schema validation, image resolution checks, document completeness verification, and duplicate detection before anything enters the pipeline.

MODEL_EXEC
Execute

YOLOv8 for vision, XGBoost for risk, LLM for documents — the right model runs for the input type.

CONFIDENCE
Score

A calibrated enterprise confidence score (0.0–1.0) — not a raw model output — weighing model confidence, historical accuracy, and context (time, phase, weather).

DECISION
Decide

Maps score to risk class: LOW (log only), MEDIUM (dashboard flag), HIGH (notify safety officer + retrieve protocol), CRITICAL (executive escalation + possible work-stoppage).

ACTION
Escalate & Mitigate

Role-based escalation routes to the right person; RAG retrieves the specific applicable protocol via Groq LLM for an actionable mitigation — not a generic rule.

LOGGED
Close

The full reasoning chain — detection through resolution — is preserved in the PostgreSQL audit log with timestamps, responsible parties, and outcome.

Accident prevention through historical learning

Stopping the next incident, based on the last one.

ACIP doesn't just detect current hazards — it learns from every project. Patterns invisible in spreadsheet reviews are surfaced by the model and applied predictively to new sites.

📊

Portfolio pattern mining

XGBoost finds multi-factor correlations humans miss — e.g. "new subcontractor mobilisation + concrete pour + temperature above 90°F" correlating with a 3.2× rise in heat incidents.

🎯

Pre-construction profiling

A new 40-story tower in Miami automatically inherits risk patterns from prior high-rise projects in hot climates — seasonal curves, trade violation profiles, critical-activity windows.

🔁

Continuous refinement

A pattern detected on Project A — scaffold violations spiking in week 3 of envelope work — is encoded into the model and applied predictively on Project B at the same phase.

⚠️

Near-miss amplification

Near-misses reveal failure modes without injury. ACIP treats them with the same rigour as incidents, vectorising the narratives for semantic retrieval when similar conditions recur.

🌦️

Environmental correlation

Weather, temperature, and wind are correlated with incident timing so risk scores rise before conditions turn dangerous, not after.

🗄️

Persistent memory

Trained models and the vector store keep safety knowledge after a project ends — institutional memory that no longer walks out the door with the team.

Beyond what a camera can see

Visual prediction — and under-land prediction.

Cameras catch what's visible. ACIP's Document Intelligence infers what's hidden below ground — utilities, soil conditions, abandoned infrastructure — by correlating decades of plans, surveys, and reports before a single shovel breaks ground.

Visual prediction

What ACIP sees

  • PPE compliance — helmets, vests, glasses, gloves, harnesses, boots
  • Unsafe conditions — unguarded openings, missing guardrails, unsafe scaffolds
  • Zone violations — crane swing radius, active vehicle routes, overhead work
  • Equipment status and per-zone headcount for mustering and overcrowding
YOLOv8n (6 MB, CPU-capable) fine-tuned on construction data — standard YOLO doesn't know what a fall-protection harness looks like.
Non-visual prediction

What ACIP infers below ground

  • Underground utilities — from as-builts, municipal maps, geotech logs, GPR reports
  • Hidden structural conditions — corroded rebar, abandoned post-tension cables
  • Environmental hazards — asbestos, lead, contaminated soil probability
  • Foundation risk — settlement, heave, lateral pressure from adjacent structures
Every inference carries a confidence score and recommends physical verification when below threshold. ACIP supplements — never replaces — investigation.
Worked example · Subsurface detection

Detecting an abandoned water line in a 60-year-old plan

A renovation in a 1960s office building needs new basement foundation work. The contractor uploads the current demolition plan. ACIP processes it and also retrieves the original 1962 mechanical drawings stored in ChromaDB at project setup. Those drawings show a 6-inch cast-iron domestic water line running diagonally through the zone now marked for excavation — capped during a 1990s renovation but never removed, so it appears on no current survey.

⚠ CRITICAL — Potential abandoned water line in excavation zone.
Source: 1962 mechanical plan, Sheet M-3, grid B4–C6. Shown as 6-inch CI domestic water. Not present on current utility survey (2024). Recommend potholing at grid B5 before excavation. Confidence: 0.72 (source >50 years old, no corroborating survey).
Technology

An enterprise AI stack, end to end.

ACIP unifies modern computer vision, gradient-boosted ML, retrieval-augmented generation, and multi-agent orchestration — every layer chosen for production reliability and auditability.

GenAI & Agent Orchestration

LangGraphLangChainGroq · Llama 3.3 70BChromaDBRAG pipeline

Computer Vision & Machine Learning

YOLOv8XGBoostPyTorchOpenCVscikit-learnsentence-transformersNumPy · Pandas

Backend

Python 3.13FastAPIPydantic v2SQLAlchemy 2AlembicUvicorn

Frontend

React 18TypeScript 5MUI v5RechartsReact Router 6Vite

Infrastructure & MLOps

PostgreSQLDockerMLflowGitHub ActionsCloudflare R2Supabase Auth
Inside the engine

LangGraph orchestration, not a linear pipeline.

Construction risk decisions branch. Should we retrieve safety protocols? Should we escalate? That requires a graph, not a chain. LRIE is built as a LangGraph StateGraph — each node a pure, typed, logged function.

vision_node risk_scoring_node should_run_rag? rag_agent_node should_escalate? escalation_node END
LangGraph

Deterministic orchestration

A StateGraph lets LRIE branch, merge, and retry without losing state. Conditional edges decide whether to retrieve protocols (risk > 0.6) and whether to escalate (risk > 0.8 or critical violation). Every node's I/O is typed with Pydantic, and the full execution trace is logged.

LangChain

The RAG pipeline

LangChain loads documents, splits them with RecursiveCharacterTextSplitter, embeds chunks via all-MiniLM-L6-v2 into ChromaDB, and runs the retrieval chain that feeds the Groq LLM. Prompt templates keep every interaction structured and consistent.

Enterprise-ready

Built for audit, compliance, and integration.

📋

Full audit trail

Every decision logged with input provenance, processing chain, model metadata, confidence rationale, action taken, and resolution — designed for OSHA, insurance, and litigation defence.

👥

Role-based escalation

Configurable hierarchies from site worker to CXO. Unacknowledged HIGH alerts auto-escalate after a configurable window; CRITICAL findings bypass the queue.

🔌

Ecosystem integration

Box, SharePoint, Procore, Autodesk Construction Cloud, Revit/Navisworks, Teams, Slack, and IoT sensor feeds — ACIP fits existing construction stacks.

🔐

Security & access

JWT auth via Supabase, role-based access at site/project/portfolio level, TLS in transit, encryption at rest, configurable data residency.

Regulatory alignment

Designed to support OSHA 1926 documentation, ISO 45001 safety-management requirements, and client-specific EHS reporting.

📈

Executive intelligence

CXO-ready dashboards with trend analysis, site comparison, and audit-ready decision trails — built on React, MUI, and Recharts.

Development status

In active development — built in the open, with discipline.

ACIP is being engineered progressively through a structured curriculum, from foundational architecture to a deployed, industry-ready platform. We're transparent about where it stands.

ArchitectureComplete — system design, LRIE state machine, LangGraph orchestration, API contracts, folder structure
FrontendInitialised — React + TypeScript + MUI; nine screens designed (Executive Dashboard, Live Risk Monitor, Vision AI, Document Intelligence, and more)
In developmentLRIE state modelling, decision & confidence engines
RoadmapML risk scoring → computer vision → document intelligence → agent orchestration → full LRIE assembly → production deployment
Deployment targetContainerised (Docker), CI/CD via GitHub Actions, cloud-native

Bring incident prevention to your projects.

We're partnering selectively with forward-looking construction teams and insurers to shape ACIP around real jobsite needs. Request a briefing.

Request a briefing