Every source, fused into one knowledge map.
CLERINT Fusion pulls all your intelligence — GEOINT, SIGINT, HUMINT, financial, open-source and broadcast — into a single knowledge graph. It resolves entities, reads the patterns, scores the threat, predicts the next move, and shows your team exactly where the gaps are — fully on-prem, or fully air-gapped.
Everything an analyst opens in a day.
Fusion is not one screen with a graph on it. These are the surfaces the work actually moves through.
- Overview
- Dashboard, temporal view, entity browser and the geographic picture.
- Intelligence
- Search, Ask Intel natural-language query, persons, events, organisations and locations.
- Threat Assessment
- Intelligence intake, submissions, assessments and the situation room.
- Analysis
- Nexus link analysis, hotspots, threats and predictions.
- Visual Intelligence
- Photo browser, face stacks, reverse search, watchlist, live alerts, review queue, cameras, generative variants and the audit trail.
- Operations
- Reports, watchlists, data import, pending review, SITREP inbox and tasking.
One knowledge graph for every entity.
Every entity and relationship the system has ever seen — resolved into a single graph. CLERINT finds the shortest path between any two people, surfaces the broker who connects otherwise-separate groups, and detects the operating cell hiding in the noise.
The network, and how it was worked out.
Every answer is reproducible and shows its working. A broker the system surfaces is a measured result an analyst can defend in a review, not a hunch about who looks important on the canvas.
Every person of interest, scored 0–100.
Threat scores aren't opinions. CLERINT computes them from seven weighted, transparent factors — the factors, the weights and the evidence trail are visible to the analyst and to oversight, every score traceable to its source.
Every report carries how much it can be trusted.
Intelligence is graded on entry using the Admiralty convention — source reliability and information confidence recorded separately, because a reliable source can still pass on a doubtful claim.
- Source reliability
- How dependable this source has proven over time.
- Information confidence
- How well this particular claim is corroborated.
- Urgency and status
- What has to move now, and what is still being worked.
- Entity extraction
- People, places, numbers and organisations pulled from the narrative on submission.
- Pattern matching
- The new report checked against everything already held, before it is filed.
- Claim tracking
- Individual assertions carry their own status, so one wrong detail does not void a report.
It shows your team where the gaps are.
CLERINT flags the missing links — the communication, the transaction, the movement that should be there but isn't — and turns each gap into a concrete collection task, directing your team to close it.
Predictions that are graded afterwards.
Every projection is scored against what actually happened — accurate, partially accurate, inaccurate. A model that is never marked is a model nobody should act on.
- Hotspots
- Geographic clusters found from the incident data itself, each labelled growing, stable or shrinking.
- Communication spikes
- A network talking far more than its own baseline.
- Financial anomalies
- Movement of value that does not fit the established pattern.
- Movement anomalies
- Travel and sightings that break a subject's routine.
- Network anomalies
- Structural change — a new broker, a severed link, a merged cell.
- Temporal anomalies
- Activity at hours or intervals that do not fit.
- Recidivism
- Likelihood a known subject re-offends, and in what window.
- Network reactivation
- A dormant group resuming contact.
- Spatial & temporal forecasting
- Where and when the next incident is most likely.
Offline map tiles ship with the deployment, so hotspots and geography work with no outbound network.
See the next move before it happens.
Dormant networks resuming contact, patterns repeating, thresholds crossed — the intelligence layer reads them and projects the likely next event with a confidence score, while there's still time to act.
Photographs, treated as intelligence.
Imagery from seizures, cameras and field collection becomes searchable, clustered and attributable — with a human in the loop before anything is asserted.
- Photo browser
- Every image held, filterable by source, case and detection.
- Face stacks
- The same person's appearances grouped across the whole collection.
- Reverse face search
- Find a face across the archive from a single photograph.
- Watchlist
- Faces of interest matched as new imagery arrives.
- Live alerts
- Camera matches raised the moment they happen.
- Review queue
- Every uncertain match waits for a human decision.
- Camera registry
- Which camera, where, and covering what.
- Generative variants
- Age, pose and appearance variants to widen a search on a poor source image.
- Audit trail
- Who searched for whom, when, and what they were shown.
It reads the systems you already run.
Fusion reads the systems you already run — no export ritual, no bespoke integration project.
Platform names and logos are trademarks of their respective owners. Their presence here identifies the sources CLERINT can lawfully collect from, and does not imply any partnership, endorsement or affiliation.
The things people ask before a briefing.
Can it work fully offline or air-gapped?
Yes. Fusion deploys on-premise or fully air-gapped, with local AI and offline maps, so hotspots, network analysis and briefing generation all work with no outbound connection at all.
Can it import call-detail records and existing case files?
Yes. CSV, JSON, Excel and TSV files, plus live connectors to the major relational and document databases, REST endpoints and existing case-record systems, with identity and phone numbers normalised on the way in.
How is the threat score calculated?
From seven weighted factors — criminal history, network position, activity recency, communication patterns, financial indicators, geographic risk and intelligence flags. The weights are configurable and visible, and every score traces to the evidence behind it.
Are the predictions checked afterwards?
Yes. Every prediction is graded against what actually happened as accurate, partially accurate or inaccurate, and that feedback is retained. A model nobody marks is a model nobody should act on.
Fuse your sources. Off the grid.
Bring your own data to a scoped, air-gapped pilot — we'll resolve it into one graph and show you the network, the threats and the gaps, without a single byte leaving your perimeter.