htm-monitor provides rare, actionable early warnings for critical systems. Unsupervised. Domain-agnostic. Validated against real historical events with known outcomes.
htm-monitor works on any system that produces continuous numeric data over time. The same unsupervised engine adapts to each domain.
Real-time anomaly detection for electrical grid infrastructure. Validated against ERCOT, CAISO, and NYISO with 6+ years of public data.
Early detection of patient deterioration from continuous vital signs. Validated on 99 real ICU stays from MIMIC-IV — predicted emergency interventions 2–12 hours ahead.
htm-monitor has been tested against real historical events with known outcomes. Here's what it found.
Real-time anomaly detection for electrical grid infrastructure
| Event | Outcome | Lead / Lag | Alert Fired | Event Onset |
|---|---|---|---|---|
| Winter Storm Uri Feb 2021 · 4.5M customers lost power |
Predicted | 98 h lead | 2021-02-11 00:00 | 2021-02-15 02:00 |
| Winter Storm Elliott Dec 2022 · Contextual alert |
Explained | — | 2022-12-23 08:00 | 2022-12-23 08:00 |
| Event | Outcome | Lead / Lag | Alert Fired | Event Onset |
|---|---|---|---|---|
| August 2020 Rolling Blackouts 813K customers · First CA blackouts since 2001 |
Detected | 6 h lag | 2020-08-15 00:00 | 2020-08-14 18:00 |
| September 2020 Follow-on Continued stress on grid |
Detected | 13 h lag | 2020-09-07 01:00 | 2020-09-06 12:00 |
| September 2022 Record Demand 51,426 MW peak · Near-miss |
Predicted | 129 h lead | 2022-09-01 08:00 | 2022-09-06 17:00 |
| Event | Outcome | Lead / Lag | Alert Fired | Event Onset |
|---|---|---|---|---|
| June 2020 Dispatch Anomaly Demand-generation imbalance |
Predicted | 45 h lead | 2020-06-22 21:00 | 2020-06-24 18:00 |
| June 2024 Heat Wave Extreme demand event |
Predicted | 69 h lead | 2024-06-19 03:00 | 2024-06-22 00:00 |
| Winter Storm Fern Jan 2025 · Gas-price leading indicator |
Predicted | 47 h lead | 2025-01-23 01:00 | 2025-01-25 00:00 |
| January 2026 Cold Snap Severe cold weather event |
Detected | 368 h lag | 2026-02-10 08:00 | 2026-01-26 00:00 |
Early detection of patient deterioration from continuous vital signs
| Patient / Event | Outcome | Lead / Lag | Alert Fired | Event Onset |
|---|---|---|---|---|
| Multi-Organ Crisis — Ventilation + Vasopressor Patient died · Intubation + norepinephrine + vasopressin |
Predicted | 20 h lead | 2024-05-16 21:00 | 2024-05-17 17:20 |
| Multi-Organ Crisis — Ventilation + Vasopressor Patient died · MI & shock · Intubation + norepinephrine + vasopressin |
Predicted | 15 h lead | 2024-10-12 01:00 | 2024-10-12 15:56 |
| Vasopressor Initiation (Norepinephrine) Patient died · Hemodynamic collapse |
Predicted | 2 h lead | 2024-02-29 21:00 | 2024-02-29 23:05 |
| Emergency Mechanical Ventilation Patient died · Respiratory failure requiring intubation |
Predicted | 2 h lead | 2024-09-15 03:00 | 2024-09-15 05:02 |
| Emergency Intubation + Vasopressor Patient died · Ventilation + norepinephrine · Sepsis |
Detected | 2 h lag | 2024-05-18 16:00 | 2024-05-18 14:00 |
| Multi-Organ Crisis — Ventilation + Vasopressor Patient died · Norepinephrine + intubation |
Detected | 7 h lag | 2024-11-09 19:00 | 2024-11-09 11:42 |
The engine is Hierarchical Temporal Memory — a biologically-inspired sequence-learning algorithm extended with a grouped-consensus decision layer.
Each signal model learns its temporal patterns continuously, adapting to drift without retraining. No fixed thresholds. No supervised labels required.
System alerts fire only when multiple independent signal models agree. This cross-group consensus layer dramatically reduces false positives.
Every alert is traceable to specific signal anomalies at specific timesteps. Per-model, per-group visibility lets operators understand exactly what triggered and why.
Runs single-process in Python. No GPU. No external API calls. Deploys on a $5/mo VPS or inside an air-gapped VM — operator's choice.
Most anomaly detection systems generate noise. htm-monitor is built for the opposite: say nothing until something genuinely matters, then say it early enough to act. Foresight is power.
We believe the same unsupervised temporal learning engine can serve any domain that produces continuous numeric data — power grids, patient monitors, industrial sensors, financial feeds. The pattern is universal: learn what normal looks like, and flag when reality departs from the learned envelope.
A system that cries wolf loses trust. htm-monitor is designed for rare, high-confidence alerts — not dashboard noise.
No labeled training data required. The system learns from the structure of normal operations and flags departures automatically.
Every alert is fully auditable, traceable to specific signals and timesteps. Operators can see exactly what the system sees.
The same engine works across power grids, patient monitoring, and beyond. If it's continuous numeric data over time, htm-monitor can learn it.
htm-monitor is developed and operated by DataGrip LLC. We're focused on making unsupervised temporal anomaly detection practical and accessible for critical infrastructure operators.
Whether you want a free technical preview, a paid pilot, or just want to learn more — we'd love to hear from you.
Send a 90-day telemetry sample and one past event you'd like scored. You get a preview audit within a week: lead/lag timing, false-positive rate, per-signal analysis. No commitment.
Full historical replay against your event catalog, with a written per-event audit. The fee credits in full against your first subscription month.