Skip to content
Intelligence

Findings that arrivewith their evidence attached

Analysis produces signals, signals become investigations, investigations become reports — and at every step the underlying records travel with the conclusion. A finding you cannot open is a finding you have to take on trust, which is the one thing an analyst should never have to do.

The DjiniousData Analyze view, filtered to locomotive G1206-AR1042: five findings — three critical — each with its type, severity, confidence, method and an Inspect action, above a severity summary of the whole fleet.The DjiniousData Analyze view, filtered to locomotive G1206-AR1042: five findings — three critical — each with its type, severity, confidence, method and an Inspect action, above a severity summary of the whole fleet.
Twelve findings over a fortnight of fleet telemetry, filtered here to one locomotive. Severity is derived from confidence rather than assigned, and every row opens onto the readings behind it.Seeded development stack · AcmeRail fixture
What it does

The working surface, part by part

  • Signals

    anomaly · pattern · weak signal

    Each signal names its subject, the method that produced it, a confidence, and the knowledge objects it was computed from. Severity is derived from confidence rather than assigned, so two findings at the same confidence cannot be ranked differently by whoever wrote them.

  • Search across everything ingested

    keyword · semantic · hybrid

    One search over entities, documents, events and insights, in three modes. Hybrid combines the full-text index with vector similarity; results carry their source and their score, and every hit opens onto the record behind it.

  • Investigations

    cases, items and provenance

    A case gathers signals, notes and evidence into a thread with an owner, a priority and a status. Evidence items keep their references back to the source objects, so a case reviewed six months later still answers 'where did this come from'.

  • Reports

    typed sections, with citations

    Structured reports whose sections are typed — narrative, entities, signals — and whose citations point at records rather than at other prose. A report is generated from the platform's state, so a figure in it and the dashboard tile it came from cannot disagree.

  • Dashboards

    aggregates computed on read

    Metric, breakdown, trend and map widgets, each a query over the ingested objects with its own facet, filter and aggregation. Nothing is precomputed, so nothing goes stale, and a tile's number is always one click from the rows that produced it.

  • Alert rules

    max · min · avg · count, per group

    A rule watches one field of one source, grouped — per unit, per site, per device — and fires when the aggregate crosses a threshold. Rules are owner-scoped: a rule evaluates against exactly what its owner is allowed to see.

The rule this page is built on

Every figure on this site was measured, including the ones in the screenshots

The findings in the captures below are not written copy. The seed that builds the AcmeRail fixture reads the ingested telemetry, computes the fleet's actual engine hours, peak coolant temperatures and minimum bus voltages, and composes the finding text from those numbers. If the fixture stopped producing the anomaly, the finding would change with it rather than becoming a lie the site keeps telling.

An open investigation case for locomotive G1206-AR1042 consolidating its cooling and electrical risk signals, with three collected evidence items, its tags, its owner and its activity trail.An open investigation case for locomotive G1206-AR1042 consolidating its cooling and electrical risk signals, with three collected evidence items, its tags, its owner and its activity trail.
The case that consolidates the fleet's two concurrent health flags on one unit. Its evidence items keep their references to the telemetry rows the flags were computed from.Seeded development stack · AcmeRail fixture
  • Signal summaries interpolate measured aggregates, never hard-coded figures
  • Each signal carries references to the knowledge objects it was drawn from
  • The alert thresholds are set against the measured fleet, so the seeded stack ships with real firings rather than four rules that have never fired
  • One locomotive breaches both the thermal and the electrical rule — the same unit the investigation case and the deep-dive report are about

The fixture is fiction — AcmeRail, its eight locomotives and its two vendors are invented — but the arithmetic is not. The figures in every finding are read from the ingested data at seed time.

From finding to action

A rule that fires is the start of the thread, not the end of it

An alert without a case is a notification somebody eventually stops reading. The platform's path runs the other way: the rule fires, the signal explains what was measured and what it implies, the case collects the evidence and assigns an owner, and the report is what leaves the building.

  • Alert history keeps the observed value and the threshold beside each firing, so a rule that is too sensitive is visible as such
  • Cases carry collaborators, tags, comments and an activity trail
  • Reports can be generated from a case, and keep the link back to it
  • Everything is org-scoped: one customer's findings are not another's to read
The DjiniousData alert history listing eleven open alerts raised by four fleet rules, each naming the locomotive, the observed value and the threshold it crossed.The DjiniousData alert history listing eleven open alerts raised by four fleet rules, each naming the locomotive, the observed value and the threshold it crossed.
Every firing keeps the value it observed beside the threshold it crossed, so a rule that is too sensitive is visible as such rather than argued about.Seeded development stack · AcmeRail fixture
Where the findings come from

A report is a view of the platform's state, not a document beside it

A report's sections are typed and its citations point at records. That is what makes the figure in a report and the tile it came from unable to disagree: they are the same query, asked twice.

A generated AcmeRail fleet maintenance and health briefing, with an executive summary, per-unit fleet status, health flags and recommended actions, each section citing the telemetry it draws on.A generated AcmeRail fleet maintenance and health briefing, with an executive summary, per-unit fleet status, health flags and recommended actions, each section citing the telemetry it draws on.
The briefing quotes distance, engine hours, peak coolant and minimum bus voltage per unit — all read from the ingested telemetry rather than written into the prose.Seeded development stack · AcmeRail fixture

Bring a question you already argue about

The best demo of an analysis surface is a disagreement it can settle. Bring the number two teams report differently, and we will show you where each one comes from.

DjiniousDataIngestion · Ontology · Knowledge graph · Agents

Ingest what you already have, model it as your own domain rather than ours, and keep every conclusion attached to the rows it was computed from.

  1. Ingest
  2. Model
  3. Resolve
  4. Act

Every screenshot on this site is a capture of the running platform.