Early Access

Picasso
Visual Operations for DriftMind

Build forecaster hierarchies of arbitrary depth, navigate from a plant overview down to a single sensor, and replay historical anomalies and Echo patterns of interest — all backed by the time-series database you already run.

Why Forecasters Need a UI

Lists Don't Scale

Once you have hundreds of forecasters across dozens of sites, a flat list is useless. You need to think in plants, lines, cells, and sensors — not in UUIDs.

Anomalies Need Context

An anomaly score on its own is just a number. To act on it, an operator needs to scrub the timeline, see the surrounding signal, and compare it against the same hour last week.

POIs Need to Be Navigable

Echo finds matches in real time. But the value of a pattern library is in being able to replay every historical occurrence, group them by severity, and learn from the false positives.

N-Level Forecaster Hierarchies

Model your operation, not your database

Picasso lets you organise forecasters into a tree of arbitrary depth — region, plant, line, machine, sensor — whatever matches how your operations team actually thinks.

Each node aggregates the anomaly scores and active Echo patterns of its children, so a site lead can spot which plant needs attention without drilling into every sensor. Drill down only when something demands a closer look.

Hierarchies are stored independently of the forecasters themselves — reorganise without data loss, and bind the same forecaster into multiple hierarchies (operations, maintenance, energy) if it serves more than one team.

▼ Madrid Plant RO Train 1 Membrane Bank A pump-1 0.04 pump-2 0.61 pump-3 0.08 ▶ Membrane Bank B RO Train 2 Energy Recovery ▶ Algeciras Plant

Anomaly Replay at Any Level

Scrub through history, not through alerts

Pick any forecaster — or any node in the hierarchy — and pull up every period where the anomaly score crossed your threshold. Picasso reads from your time-series database to reconstruct the exact signal that triggered each event.

Compare the same signal across days, weeks, or months. See the leading indicators in other channels. Decide whether the anomaly was real, a sensor fault, or an operational change you didn't account for — without writing a single SQL query.

Mark events as true positive, false positive, or operational. Picasso uses the labels to surface drift in the underlying data and to suggest threshold tuning per forecaster.

Anomaly Events — pump-2 (last 30 days) 2026-04-22 14:31 score 0.78 CRITICAL duration 18 min · vibration leading [ ✓ true positive — bearing replaced ] 2026-04-19 09:12 score 0.71 CRITICAL duration 4 min · pressure transient [ ⊘ operational — backwash cycle ] 2026-04-15 03:44 score 0.64 MAJOR duration 11 min · temperature drift [ ✗ false positive — sensor fault ]

Echo Patterns of Interest, Replayable

Every match, indexed and navigable

For every Echo pattern attached to a forecaster, Picasso keeps a searchable history of matches. Filter by severity, by timestamp, by pattern, or by hierarchy node. See the actual signal that triggered each match overlaid on the reference pattern.

This is where pattern libraries become useful instead of theoretical. When a new failure mode appears, browse the existing patterns to find the closest historical match. When you create a new pattern, immediately replay it against the last 90 days of data to see whether it would have caught the events you care about.

Patterns of interest become a shared institutional memory — fault signatures captured once, available to every operator forever.

Pattern: bearing-failure (CRITICAL) Matches across hierarchy — last 90 days Madrid · RO 1 · pump-2 3 matches latest: 2026-04-22 — score 0.92 Madrid · RO 2 · pump-5 1 match latest: 2026-04-08 — score 0.88 Algeciras · RO 1 · pump-1 2 matches latest: 2026-03-30 — score 0.81 → click any match to replay against the reference signal

Backed by the TSDB You Already Run

Picasso doesn't store time-series data itself — it queries the database you already use. Anomaly events and pattern matches are indexed by timestamp and forecaster, then resolved against your TSDB on demand.

InfluxDB
TimescaleDB
QuestDB
Prometheus
VictoriaMetrics
AVEVA PI
ClickHouse
Custom REST

Don't see yours? Picasso ships with a generic REST adapter — point it at any endpoint that returns timestamped values for a given tag.

Built for Operations Teams

Plant Operations

Industrial Sites

Group forecasters by line, machine, and sensor. Operators watch a single dashboard; engineers drill into anomalies with full historical context. Echo pattern libraries become shared maintenance knowledge.

Network Operations

Telecom NOCs

Build hierarchies that mirror your network topology — region, market, cell site, sector. Replay degradation events alongside the KPIs that triggered them. Pattern libraries for known failure signatures.

Data Centre Operations

Data Centres

Organise forecasters by data hall, row, rack, and unit. Replay thermal events to understand cascade behaviour. Build pattern libraries for cooling failures, power anomalies, and PUE excursions.

Get Early Access to Picasso

Picasso is in early access for DriftMind customers and partners. Join the program to shape the product, get priority onboarding when it ships, and lock in early-access pricing.

Request Early Access Learn About DriftMind