_{PROPOSAL · FOR NIUM · AVI REDDY}

Money Movement 2.0 — High-Level Design

Scope. A payout control plane that sits beside Nium's existing payout pipeline and turns six new flow types — stablecoin settlement, marketplace split payouts, high-risk verticals, agent settlement, prediction-market resolution payouts, perpetuals collateral/PnL — into policies on one rail instead of six products. Built from Nium's public documentation and job postings only; every integration point below is a documented Nium API or a named public partner.

1. Design goals and non-goals

Non-goals: acquiring; issuing a stablecoin; replacing Transfer Money, Wallets, Onboarding or RFI; custody of digital assets (Nium's stated model is "Nium does not hold USDC").

1a. Design doctrine — six rules the plane is built by

2. Context — where it sits

flowchart LR
  subgraph Clients
    MK[Marketplaces & platforms]
    TR[Corporate treasuries]
    VN[Venues: prediction, perps, exchanges]
    OP[Licensed operators]
    AG[Agents / facilitators x402·MPP·AP2]
  end
  subgraph MM2[Money Movement 2.0 — control plane]
    API[Control-plane API<br/>POST /payouts · GET /trace]
    POL[Policy service]
    EV[Evidence gate]
    RT[Appetite router]
    ST[Settlement orchestrator<br/>timing · netting · rail]
    LG[Ledger & reserves]
    TRC[Trace & audit store]
  end
  subgraph Nium[Nium platform — documented surface]
    TM[Transfer Money / Batch / Validator / Rail preference]
    WL[Wallets & multi-currency accounts]
    ON[Customer Onboarding v5 · RFI v5]
    VF[Nium Verify]
    FW[Fund Wallet · VANs · USDC funding]
    CD[Cards · Thredd · stablecoin cards]
    KF[(Kafka · webhooks)]
    RS[(Redshift · dbt)]
  end
  subgraph Partners
    CB[Coinbase · Circle CPN]
    BK[Bank & rail partners<br/>JPM · DBS · SCB · Banking Circle · ClearBank · Column · CFSB · Cross River · Cuscal · Pateno · Cobre · Netbank · GMO · BS2]
    RG[Licence registries · venue attestations · oracles]
  end
  Clients --> API
  KF --> API
  API --> POL --> EV --> RT --> ST --> LG --> TRC
  EV --> ON
  EV --> VF
  EV --> RG
  ST --> TM
  ST --> FW
  ST --> CD
  FW --> CB
  TM --> BK
  TRC --> RS
  TRC --> KF

3. Components

4. The five decisions (per payout)

flowchart TD
  A[Payout intent] --> S{1 · Source<br/>asset allowed?<br/>convert JIT / prefunded}
  S -->|not allowed| D1[DECLINED · SOURCE_ASSET_NOT_ALLOWED]
  S --> E{3 · Evidence<br/>providers in parallel<br/>fold: PASS · HOLD · DECLINE}
  E -->|DECLINE| D2[DECLINED · reason code]
  E --> R{2 · Entity<br/>appetite router}
  R -->|no pairing| D3[DECLINED · JURISDICTION / REGISTRATION / VERTICAL]
  R -->|above max ticket| H1[HELD · TICKET_ABOVE_MAX · EDD case]
  E -->|HOLD| H2[HELD · RFI / pending]
  R --> T{4 · Timing<br/>now · JIT · NET · EVENT}
  T -->|event missing| H3[HELD · AWAITING_EVENT]
  T -->|window open| Q[QUEUED · NETTING_QUEUED]
  T --> L{5 · Rail<br/>choice ∩ policy ∩ partner<br/>caps · 24/7 windows · preference}
  L -->|none| D4[DECLINED · RAIL_UNAVAILABLE]
  L --> P[Price: fee · reserve · conversion · FX at the net · rail fee]
  P -->|rail fees > 25% of value| H4[HELD · UNECONOMIC_ROUTE → fallback to netting]
  P --> G[Ledger postings · balanced per currency]
  G --> OK[APPROVED · trace]

5. Money-flow models the plane runs

1. Pre-funded fiat payout — client wallet (safeguarded EMI money / US sponsor-bank custodial account) → Nium nostro at the destination partner → local rail.
2. Just-in-time stablecoin funding — USDC to a static address → Circle/Coinbase custody + conversion → USD wallet → payout. Nium never holds the asset.
3. Marketplace three-tier split — PSP sweep → Platform Receivables → Fees / Seller MCY / Base → seller payout by seller choice (bank, USDC, card).
4. Event-triggered payout — resolution / delivery / attestation arms the payout; evidence gate; immediate rail.
5. Streamed-then-netted — micro-receipts in escrow → one rail-sized payout per beneficiary per window (or threshold), FX at the net.
6. Collateral on-ramp / PnL off-ramp — local RTP collection → USDC → venue; USDC PnL → local fiat, with venue attestation.
7. Risk-reserved payout — fee + rolling reserve + partner-scoped appetite; sender pays on top; reserve released T+N.
8. Scheme-level stablecoin settlement of Nium's own card book (SoFi/Mastercard pattern, Visa USDC settlement pilot) — a policy on the settlement asset, not a product.

6. Non-functional requirements

7. Deployment on Nium's documented stack

Java/Spring Boot services on Kubernetes (AWS), MySQL for policy/appetite/trace metadata, Kafka for intents and trace events, Redis for evidence TTL cache and idempotency, S3 for trace archives, Redshift/dbt for reporting, Datadog/Prometheus for SLOs, Terraform, Jenkins/ArgoCD. The reference implementation here is dependency-free JavaScript so the same engine runs in tests, a CLI, an HTTP server and the browser demo; the production port is a one-to-one translation of src/engine/* into Spring services.

8. Rollout (matches the strategy's four quarters)

Q4 2026 — policy engine v1, trace on the existing payout path (shadow mode: decide, don't route), appetite router with three partners, stablecoin payout beta. Q1 2027 — marketplace seller-choice + lineage; two prediction-market pilots (US DCM B2B, LatAm). Q2 2027 — HRV GA with priced verticals and reserves; perps on/off-ramp; margin-time SLA. Q3 2027 — agent registry + facilitator + netting GA.

_{Money Movement 2.0 · an independent proposal by Avi Reddy for Nium. Not a Nium document. Confidential, shared for evaluation.}