What you're looking at
Your data centre sensors already collect continuous temperature, humidity, and power telemetry.
In traditional monitoring, alerts often trigger only when one channel crosses a static limit.
The problem: by the time a single reading breaches a threshold,
the fault often has been developing for a long period. Many failures drift gradually
while individual values still appear "within limits."
What Sentinel does: instead of watching individual readings, it learns
the normal pattern across all available sensor channels together. It understands that when
intake temperature rises, exhaust should rise proportionally. When one power phase
creeps up while others don't, something is wrong — even though no single reading has
breached any limit.
This dashboard shows four common failure modes on synthetic (simulated) data:
- Fan degradation — gradual thermal drift as airflow reduces
- PDU overload — one power phase creeping up (asymmetry)
- Hot spot — exhaust temperature rising while intake stays normal
- HVAC drift — humidity becoming unstable as cooling degrades
For each fault, the "Lead time" shows how many hours earlier Sentinel
detects the problem compared to a traditional single-channel threshold alert.
Beyond individual fault detection, Sentinel also tracks system health over time.
Some problems don't show up as a single unusual reading — they show up as patterns. An HVAC
system that oscillates between 20°C and 24°C every ten minutes. A sensor that's been stuck on
the same reading for hours. A power supply that trips and resets every 45 minutes. Sentinel
catches these by watching the sequence of readings, not just individual values.
The "System Health Trajectory" card shows whether each fault type is healthy, slowly
degrading, or actively
pathological — and estimates how many hours until the problem becomes critical if the
current trend continues.
The bottom line
Sentinel gives you hours of advance warning before a problem
becomes an emergency. No new hardware — it works on the sensor data you already
collect. Software-only intelligence on your existing infrastructure.
The technology is sensor-agnostic — it works with any combination
of environmental, power, mechanical, or process sensors. The demo shows four
data centre fault types, but the same engine adapts to any domain where early
fault detection matters.
Your data, your asset
Every site you monitor builds a structured record of equipment behaviour —
fault signatures, degradation patterns, normal baselines. Over time, this
becomes a valuable intelligence asset. Equipment manufacturers pay for exactly
this kind of real-world reliability data but can rarely obtain it from
production environments. Your monitored estate isn't a cost — it's an
investment that grows in value with every month of operation.
How it works
Sentinel encodes multivariate sensor windows into sparse representations using a
stack of signal processing primitives developed by Sparse Supernova.
1. Windowing & Encoding
Multivariate sensor windows are encoded into a compact representation that captures
cross-channel behavior, then normalized against baseline operating behavior.
2. USL Alignment
A USL-aligned sparse representation is used to keep model capacity matched to
observed signal complexity, reducing overfit and preserving deterministic behavior.
3. Sparse Feature Extraction
A sparse encoder learns the joint distribution of normal behavior.
Conformal gating limits update pressure on already well-modeled patterns,
improving stability and compute efficiency.
4. Anomaly Detection
At inference, each window is scored with conformal thresholding and sustained
anomaly confirmation logic. This reduces noise-triggered alerts while preserving
early warning on developing faults.
5. Sequence Pathology (USPD v2)
While USAD scores individual windows, USPD (Universal Sequence Pathology
Detector) analyses the sequence of classifications over hours or days. It detects
chronic pathologies invisible to per-window scoring:
- Oscillation — system cycling between states (high ordering defect)
- Stuck sensor — no state variation (low coverage health)
- Intermittent faults — periodic state cycling across multiple scales
- Degradation trajectory — sustained worsening over time
Sequence-level scoring uses calibrated trend and ordering signals to classify
broad trajectory states (healthy, slowly degrading, degrading, pathological)
without relying on single-point threshold logic.
6. Fault Classification
A lightweight classification layer maps anomalies to operational fault classes
(fan, PDU, hotspot, HVAC), and can be tuned with site-specific incidents over time.
Primitive Stack
| USL | USL-aligned architecture sizing from observed signal complexity |
| USAD | Conformal anomaly detection — distribution-free finite-sample guarantees |
| USPD | Sequence pathology — oscillation, stuck sensors, degradation trajectory |
| SatConform | Governance envelope — monitors for representation phase transitions |
| KK V3 | Spike-ready encoding — neuromorphic deployment path via TTFS |
Data Intelligence Model
Sentinel outputs are not raw sensor data — they are structured, deterministic,
content-addressed representations (SmartAtoms) containing structured fault classifications, confidence signals, and degradation trajectories. Each
observation is governance-receipted with full provenance.
Over time, a monitored estate accumulates a dataset of labelled fault
intelligence that is difficult to obtain through any other means. Equipment
OEMs conduct accelerated life testing in labs but almost never have access to
real-world degradation trajectories from production environments at scale.
Partners retain commercial rights over insights derived from their data.
Anonymised, aggregated fault intelligence may be offered to manufacturers
and industry bodies by mutual agreement — creating a secondary revenue
stream from the monitoring infrastructure.
Operational monitoring → Structured fault intelligence → OEM product data
The same infrastructure that prevents downtime also builds a data asset that
appreciates over time.
Pure ESM · production-ready deployment path