← NEXUS Portfolio
Platform Roadmap

From one substrate to an operating platform.

A bank runs sanctions, transaction monitoring and case management as three separate systems, with an intelligence function above and investigators below. NEXUS connects them at the substrate — one resolved graph carrying identity, ownership, risk, typology and money flow, so every layer is a view on one truth rather than a system to reconcile. The detection chain and the fused investigator console are built; the arc below is the destination and where it stands today.

Where we are

Current position

The chain is built, published and reproducible.

The substrate, the detection and scoring across every crime family, and the fused investigator console are built and merged — governed, deterministic, and tested end to end. Beta v2.0 added the operations MI surface; Beta v2.1 took the copilot to three live slices, published five fused investigation cases, completed doctrine for all 57 typologies, manufactured the first capital-markets product, and made every published run reproducible from a clean clone. Since then the operations-management layer is complete — a tamper-evident case event log as the system of record, the SAR lifecycle, SLA orchestration and QA oversight — and the TM Tuning Studio has finished all ten weeks. The threat-intelligence layer now reads live feeds on a weekly schedule. In build: whole-estate batch testing. Next: the replicable investigator workspace, one object per typology.

Position · Beta v2.1 · updated 2 October 2026

The journey

Built

Intelligence foundation

A resolved-entity and typology substrate: the canonical data model, one shared entity key, and the resolved global graph. Every layer above is a view on this.

Built

Detection & scoring

Typologies across the crime families — network, trade-based money laundering, sanctions, correspondent banking and capital markets — with governed risk scoring into a ranked alert table.

Built

Fused investigation

One console where investigation and case management meet on the substrate: the graph, the case lifecycle, and an explainable, provenance-linked SAR draft — investigate and manage on one screen. Five fused investigation cases are now public, each linked to the others.

Operational surface

A replicable investigator workspace — one alert, one three-panel object (the pattern · the tunable score · the governed decision) — identical for every typology, gated by a full functional test matrix.

Built

Operations management

The supervisor's layer above execution, now complete. Management information and KPIs (Beta v2.0) — a supervisor MI dashboard and a Tableau-style MI Explorer (aging · jurisdiction · typology, with a CSV extract) — sit on a tamper-evident case event log that every action is written to. On that log: a SAR lifecycle from referral to the nominated officer through to the regulator, with its deadlines; SLA orchestration that records every breach, routes it to the supervisor and allocates waiting work by priority and capacity; and QA oversight that reviews decisions after they are made, with reviewer independence enforced. Open the ops dashboard →

Built

Detection tuning

The TM Tuning Studio stress-tests the monitoring rules against a population whose truth is known — precision, recall and investigator capacity, changed only through a governed decision. All ten weeks built; eight decisions logged — two holds on the £75k line, a two-line split for retail and everyone else awaiting the governed apply step, a customer-baseline challenger advanced to validation, a drift test that found the flat lines unstable and the customer-baseline rule stable, a ranked-score test that made the score cap, not the score, the thing to watch, a validation under independent challenge that named the two-line split as the candidate, and a capstone on a new cash rule that found capacity, not the line, decides what is caught. Open the briefing →

Intelligence layer & enterprise

On the Knowledge Centre's doctrine, the platform learns and stays current. The AI copilot has three live slices — SAR draft, EDD memo, disposition support — on grounded context; the threat-intelligence layer that pins live advisories and actors to the typology library is live — slice 1 extracts advisories, scored against a labelled set, and resolves the actors they name; and slice 2 reads OFSI, FinCEN and OFAC on a weekly schedule, with a person accepting every record. Then scale across domains. Explore the future capabilities →

Guiding principles

One substrate Identity, ownership, risk, typology, sanctions and money flow on one resolved graph — every layer a view on one truth, not a system to reconcile.
Governed & deterministic Every build is versioned, validated and byte-identical to rebuild. Nothing ships without passing the gate.
Evidence led Outputs are structured around evidence, risk, confidence, rationale and the action to take.
Tested to ship Every typology passes a five-layer functional matrix — positive and negative, on full data — before it goes live.
Tunable, not hard-coded Scoring lives in governed, versioned configuration — thresholds and weights are tuning elements, stress-tested in the Tuning Studio and calibrated on evidence.
Private IP protected Public pages show capability and architecture while the proprietary analytics and implementation stay private.

The chain is built — the work now is the operational surface.

The near-term focus is the replicable investigator workspace and rolling it across every typology, under the operations-management layer now built above it. The destination is a reusable financial-crime intelligence operating platform that connects analytics, entity resolution, investigation and governance on a single substrate.