Operational truth for critical infrastructure under stress

Evidence-native operational truth platform

Atlas4D turns fragmented sensor, weather, vision, network and operator signals into an auditable 4D evidence ledger. It separates observed reality from predictive output, measures trust mathematically, and helps operators act under uncertainty.

4.1M+
Observations
18.3M+
Evidence details
400k+
Trust metric points
4
Signed LOIs
Awesome PostgreSQL Mentioned in Awesome PyPI v0.3.0
See the platform For investors
Burgas Bay port-infrastructure pilot
EIC Accelerator 2026 applicant
Pillar 1
Operator clarity
Now
Pillar 2
Controlled actuation
Burgas Bay pilot
Pillar 3
Trust-bound copilot
Roadmap
The problem

Critical infrastructure fails at the exact moment its sensors become least reliable.

Extreme weather, congestion, equipment drift and network degradation do not only create hazards. They degrade the systems used to understand those hazards. Atlas4D adds the missing layer: exposure x trust.

01 / Degraded signals

Sensor output is not automatic truth

Fog, precipitation, vibration, packet loss and device drift can make cameras, radar, IoT and network telemetry confidently misleading.

02 / Exposure

Operators need to know what is at risk

A forecast is not enough. The operational question is which asset, zone, crew, route, service or population is exposed.

03 / Trust gate

The system must know when not to forecast

In critical operations, a confident wrong answer is worse than observed-only mode. Atlas4D makes suppression explicit and auditable.

Atlas4D ideology

Operational truth, not hallucinated prediction.

Atlas4D does not treat sensor data or AI output as automatic truth. Every observation, forecast, override and decision becomes an evidence claim with source, time, lineage and trust context.

Truth

Reality before prediction

Observed state is never silently mixed with forecast output.

Evidence

Evidence before confidence

Every important claim carries source, time, provider, context and derivation.

Review

Overrides are new claims

Human review does not erase history. It creates a linked, governed claim.

Trust

Trust is computed

Coverage, calibration, drift, lineage and review state qualify decisions.

Action

Action is bounded

The product arc is operator clarity -> controlled actuation -> trust-bound copilot.

Ontology Doctrine

A machine contract for operational truth.

Atlas4D treats ontology as an operational contract, not a decorative knowledge graph. Every object, claim, derivation, review and action must be grounded in evidence, time, trust and governance.

From observation to action
Observation → Evidence → Trust → Action

Atlas4D is built as a chain of trust: raw observations become evidence, evidence is preserved with lineage in append-only ledgers, trust is computed from agreement and uncertainty, and the resulting state can drive policy and action.

01
Evidence-native
Nothing important exists without an evidence anchor.
02
Time-native
Truth is positioned across observed, valid, system, issued and as-of time.
03
Derived-by-design
Forecasts, fusion, confirmations and decisions become first-class claims.
04
Trust-aware
Local confidence is separated from calibrated trust measured over time.
05
Governance-first
Truth is append-only, tenant-safe, reviewable and auditable.
06
Edge-sovereign
Local evidence and safety decisions can operate before cloud synchronization.
07
Action-linked
Ontology must support reviewable decisions, approvals, controls and reversals.
Mathematical Trust Engine

Models that measure whether a signal can be trusted.

Atlas4D combines operational evidence with statistical methods that evaluate agreement, uncertainty, calibration, drift and signal reliability under degraded real-world conditions.

MALP

Maximum Agreement Linear Prediction

Aligns degraded signals into calibrated operational truth by maximizing agreement with observed reality.

CCC

Lin's Concordance Correlation Coefficient

Measures agreement between prediction and observation, penalizing both bias and scale error.

Monte Carlo

Scenario uncertainty

Samples operational scenarios to expose uncertainty bands, escalation paths and what-if outcomes.

Calibration

Conformal prediction

Builds reproducible uncertainty intervals from evidence rather than assumed model confidence.

Signal correction

Kalman filtering

Continuously corrects noisy operational streams while preserving traceable state transitions.

Fusion

Anomaly detection

Detects abnormal patterns, sensor inconsistencies and behavioral deviations across evidence streams.

Weather stress
Storm, fog, precipitation and heat expose assets and degrade sensors.
Sensor degradation
Vision, radar, telemetry and network signals become noisy or incomplete.
MALP / CCC / calibration
The platform measures agreement, bias, coverage and uncertainty.
Trust gate
Atlas4D forecasts, suppresses, escalates or stays observed-only.
Operator decision
Actions remain linked to evidence, time, review and governance context.
exposure x trust
For investors and strategic partners

Decision-grade operational truth, not another dashboard

PROBLEM
Operators drown in conflicting sensor streams, stale predictions, and unverifiable alerts. Truth and prediction are confused.
ANSWER
Atlas4D separates observed reality from predictive output. Append-only ledgers, ontology-bound primitives, and chain-of-trust derivation give operators truthful state.
STATUS
Operational truth layer with 4.1M+ current canonical observations, 18.3M+ evidence details, 400k+ trust governance records, and 4 signed LOIs.
DEFENSIBILITY
Append-only evidence spine, ontology-bound primitives, calibrated trust metrics, and a deep spatiotemporal data foundation. Not more predictions, but more truthful operational state.
Read full investor brief
Architecture

Chain of trust

Operational truth in Atlas4D is a chain. Each step is a linked claim, not a rewrite. Corrections and overrides do not erase history; they create new claims that reference the prior state.

raw
→
fused
→
confirmed
→
policy decision
→
override

Five core ledgers carry these claims under database-level mutation defense, hardened by per-table block triggers and append-only role grants. Operators can correct, suppress, or override without ever rewriting historical truth.

Doctrine

Ontology-bound primitives

Atlas4D primitives are bound to schema reality. The ontology primitives matrix records consolidation status per primitive: some are canonical, some are canonical with caveat, some remain fragmented and explicitly tracked for closure.

Canonical-ready

Observation and derived observation. Single canonical surface, stable derivation taxonomy, full lineage support.

Canonical-with-caveat

Entity and review action. Canonical surfaces exist with documented edge cases for cross-source identity and write paths.

Fragmented (tracked)

Relation, decision, and trust context. Surfaces exist across multiple tables and views; consolidation is in flight, not pretended.

Capabilities

Six platform capabilities

Atlas4D capabilities are platform-grade, not feature-tour bullet points.

01

Evidence-native ingestion

Canonical evidence envelopes with provider namespaces, time, location, metric, and source context preserved through ingestion.

02

Chain-of-trust lineage

Parent and root lineage with explicit derivation semantics for fused, predicted, confirmed, and override claims.

03

Trust metrics and calibration

400k+ trust governance records, coverage semantics, drift, calibration runs, and operational trust state.

04

Scenario and threat reasoning

Forecasts, scenario branches, and threat windows that separate observed reality from predictive output and qualify forecasts with trust context.

05

Human-in-the-loop governance

Review, confirm, reject, suppress, and override workflows that never mutate historical truth. Decisions are linked claims.

06

Infrastructure telemetry

Network Guardian: 1.9M+ infrastructure telemetry metrics with SNMP/Ping collection, ML anomaly models, MVT visualization, and pgvector semantic search.

Interfaces

Operator and analyst interfaces

Live platform surfaces. Some views require authentication.

Natural language queries

Ask operational questions across time, space, and evidence in plain language. Atlas4D resolves intent against the live evidence ledger.

STSQL

Spatiotemporal SQL extension for motion, risk, forecasts, and evidence windows. Operator and analyst grade.

Evidence read APIs

Inspect recent evidence, details, and lineage through stable API surfaces with trust context attached.

Natural language queries → STSQL surface → Docs assistant →

Some surfaces require authentication.

Architecture

Built as a platform, not a vertical demo

Atlas4D is layered: an evidence and trust platform spine, operational applications on top, and vertical wedges that consume the platform without redefining it.

From signal to action
Observation → Evidence → Trust → Action

This platform spine connects sensor or human observations to evidence records, evidence to trust evaluation, and trust to operator-facing action paths without breaking lineage.

Platform spine
Evidence and trust spine. Ingestion, evidence contracts, observations, lineage, trust metrics, append-only ledgers with database-level mutation defense.
Applications
Operational applications. Mission Control, Trust Ops, Vision Evidence Ops, Network Guardian, NLQ and STSQL surfaces.
Verticals
Vertical wedges. Ports and coastal operations, telecom and infrastructure, smart city resilience, industrial operations. Wedges consume the platform; they do not redefine it.
Pilot wedge

Burgas Bay port-infrastructure pilot

Pillar 2 of the Atlas4D arc is controlled actuation. The first concrete pilot path runs through Burgas Bay port-infrastructure: cranes, sensors, weather, and operator decisions on a live coastal critical-infrastructure surface.

Why ports

Coastal port infrastructure is climate-exposed, multi-source, real-time, and operationally consequential. Operators already think in evidence and trust terms.

What Atlas4D contributes

Evidence-native fusion of radar, weather, IoT, video, and operational telemetry. Trust context attached to forecasts. Auditable operator review.

Status

Phase 0 scope: operator interviews, network and device inventory, and controlled-actuation protocol design. No autonomous control. Operator-bound throughout.

Use cases

Where Atlas4D applies

Atlas4D is built for operational environments where evidence, trust, and lineage matter more than narrative dashboards.

Ports and coastal operations

Operational truth for coastal infrastructure: weather, risk, multi-source evidence, controlled operator action.

Telecom and network operations

Network Guardian telemetry, anomaly detection, trust-aware infrastructure state across SNMP and Ping.

Smart city and municipal resilience

Event risk, weather, movement, IoT signals, operator attention surfaces with calibrated trust.

Industrial operations

Sensor fusion, evidence lineage, and controlled operational action with explicit decision history.

Auditable critical infrastructure

Auditability, provenance, lineage, and human-in-the-loop review for regulated environments.

Drone and movement analytics

Trajectory analysis, threat windows, scenario reasoning across moving platforms.

Threat forecasting → Event risk map → Investor brief →
Stack

Technical proof

Atlas4D is built on stable, audited components. Architectural choices are made for evidence integrity, lineage support, and operator-grade reliability.

PostgreSQL / PostGIS / Timescale

Spatiotemporal evidence and time-series spine. Append-only invariants enforced at the database level.

pgvector and semantic search

Evidence retrieval, document assistance, and operational query support across text, code, and structured artifacts.

Stable 64-bit IDs where used

Idempotent ingestion, dedupe registries, and controlled write paths. Stable identity where it matters; canonical UUID surfaces where it matters more.

FastAPI microservices

Gateway, monolith API, trust, vision, NLQ, NetGuard, and tile services. Containerized with Docker Compose.

MapLibre and MVT

Vector tile rendering for geospatial layers across web operator surfaces. Public landing is static HTML with lightweight JavaScript modules; this refresh does not require the React/Vite build pipeline.

Observability

Prometheus, Grafana, audit logs, trust health views, service metrics, and sentinel-based regression discipline.

For investors and strategic partners

Atlas4D is an evidence-native operational truth platform with 4 signed LOIs and a Burgas Bay port-infrastructure pilot path. Across observation, evidence, anomaly, trust, and infrastructure ledgers, the platform now manages 68M+ operational records.

Read investor brief