Loading α²Buildings…
α² Building Intelligence Lab · Private Beta

Fix the building
before anyone complains.

Predictive Comfort & Energy AI learns each zone's thermal response, CO₂/occupancy rhythm and energy signature across the KNX/BACnet/Modbus/DALI bus — then pre-empts comfort drift and trims energy without comfort loss. On-prem. EU-resident. Auditable.

How it works
18%
avg energy cut · pilot
−63%
comfort tickets
9min
median pre-fault lead
LIVE · KNX BUS · LAST 24 H
Zone monitor
ZONE-L4-07
CONFIDENCE 0.91
00:0006:0012:0018:0024:00
3 anomalies · last 24 h α²-BIM v0.7 · on-prem · 38 ms
18% energy cut · pilot predicts comfort drift per-zone model on-prem inference EU data residency KNX · BACnet · Modbus · DALI auto-tune setpoints full model card weekly auto-retrain 18% energy cut · pilot predicts comfort drift per-zone model on-prem inference EU data residency KNX · BACnet · Modbus · DALI auto-tune setpoints full model card weekly auto-retrain
Why it's different

Not another rule or schedule.

Rules and schedules trip on a fixed threshold. The Building Intelligence Lab learns each zone's signature, so it catches subtle drift early — and ignores the legitimate edge cases that flood a rule engine with noise.

Predictive, not reactive

Rules and schedules fire after a threshold is crossed. The lab models each zone's trajectory and pre-empts comfort drift before occupants ever feel it — median 9 min of pre-fault lead in pilot.

Per-zone signature

Every zone learns its own thermal response, CO₂/occupancy rhythm and energy draw across the KNX/BACnet/Modbus/DALI bus. A south-facing meeting room's "normal" is not a north-facing open-plan's — and the model knows.

Transparent & auditable

Every prediction lists the features that triggered it. On-prem / edge inference, EU-resident, with a published model card — features, residency, latency — for your ESG and audit reviewers.

By the numbers

Early pilot · 12 buildings.

Illustrative pilot & target figures from the private beta — not live production metrics.

18%
avg energy cut · pilot · 12 buildings
−63%
comfort tickets prevented · 90-day pilot
214
faults caught pre-failure · actuators · valves · sensors
38ms
median on-prem inference · edge node · EU-resident
How it works

From bus to action — sense · predict · act.

01

Sense

KNX/BACnet/Modbus/DALI telemetry streams in and is rolled up into per-zone features — temperature, CO₂, humidity, occupancy, valve/actuator state and energy draw, windowed in real time.

02

Predict

A temporal model compares the live window against the zone's own history plus occupancy and weather context, emitting a comfort-drift / anomaly score continuously — not against a one-size-fits-all rule.

03

Act

Above threshold? Auto-tune the setpoint/scene/schedule and route an alert with its escalation path. Below? Silent log + daily digest. Quiet hours and escalation are fully configurable — operators stay in control.

What the model watches

Three pattern families.

Each tuned per zone. Toggle individually. Every detection logs its confidence and the features that triggered it.

Pattern · 01

Comfort drift before complaint

Predicts CO₂ build-up, temperature and humidity drift in a zone before occupants notice it. Pre-empts the "it's stuffy in here" ticket while there's still time to re-balance the air.

Pattern · 02

Energy waste & schedule drift

Phantom loads after hours, HVAC running against an empty zone, setpoints that crept off plan. Flags the silent kWh that schedules and rules quietly leak every night.

Pattern · 03

Equipment degradation / pre-fault

Catches failing actuators, sticking valves and drifting sensors from their telemetry signature — days before they fail outright, so maintenance is planned, not reactive.

Private beta · onboarding next quarter

Apply to the
Intelligence Lab beta.

We're onboarding 12 buildings next quarter. A pilot starts on 2 zones, runs 14 days, then you decide whether to promote it estate-wide. You get a dedicated channel with our ML team and a published model card for your ESG and audit reviewers — all on-prem, EU-resident.

14-day pilot · 2 zones to start
Onboarding 12 buildings next quarter
Dedicated channel · ML team on hand
Weekly auto-retrain on your own data
On-prem / edge inference · EU data residency
Published model card for ESG & auditors