Back to VaultAI Financial VDR IP Defense

Virtual data room

Technical acquirer room for VaultAI.

This is not a login-screen demo. It is the engineering evidence a technical buyer expects before considering source-code, IP, or exclusive-rights acquisition: architecture, deterministic settlement, stress output, and migration controls.

VDR modules

10k

concurrent transaction transcript

0.000%

variance target

migration phases

Buyer diligence index

The five artifacts a bank CTO will look for first.

Each module is designed to move the buyer out of a user-demo mindset and into an acquisition-review mindset: system boundaries, mathematical invariants, measured behavior, and implementation risk.

High-Level Architecture Diagram

AWS ingress, compute, Postgres/Supabase ledger state, async Gemini compliance workers, queueing, and audit storage.

DFVL Mathematical Proof

Zero-float integer execution, atomic debit-credit equality, canonical SHA-256 settlement hashes, and PQC-ready signing posture.

Shadow Ledger Stress Evidence

Sanitized terminal transcript for 10,000 concurrent postings, sub-10ms lab p99, replay determinism, and zero variance.

Migration & Cutover Playbook

Parallel run, real-time shadow verification, fail-safe rollback, and final switch-over controls for legacy COBOL estates.

Enterprise IP Valuation Evidence

Bridge the technical VDR to the structured strategic-control thesis with replacement-cost, provenance, and diligence guardrails.

File 01

Enterprise Cloud Architecture & Data Pipelines

This is the nervous system of the software: cloud backbone, database modeling, and AI compliance paths mapped so buyers can see where latency is controlled.

Cloud backbone

Route 53

AWS WAF

ALB / API Gateway

ECS Ledger API

Public traffic is scrubbed before compute allocation. The ledger service remains region-bound and private-network aware.

Database modeling

PgBouncer

Postgres / Supabase

Journal Tables

Merkle Roots

Connection pooling bounds concurrency. Ledger tables store integer micro-units, ordered events, postings, and immutable audit roots.

AI integration

Outbox

SQS / EventBridge

Gemini Worker

Risk Findings

Gemini runs outside the posting transaction. AI latency cannot slow or corrupt deterministic settlement.

Ingress

AWS WAF
ALB / API Gateway
TLS and tenant guard

Compute

ECS ledger API
Lambda workers
Next.js dashboard

Ledger

Postgres / Supabase
PgBouncer pooling
Immutable audit tables

AI Compliance

Outbox events
Async queue
Gemini findings

Low-latency routing

The ledger API returns after deterministic validation, atomic posting, and outbox creation. Gemini review is not allowed inside the ledger write transaction.

Connection pooling

PgBouncer or Supavisor transaction pooling bounds database concurrency so traffic spikes reject cleanly instead of exhausting Postgres connections.

Async AI compliance

Outbox workers feed Gemini compliance checks through retry and DLQ paths. Findings are attached later with prompt version, model name, and SHA-256 evidence hashes.

Mermaid architecture source for buyer docs

verified exhibit

graph TD
    U["Bank channels"] --> WAF["AWS WAF"]
    WAF --> ALB["ALB / API Gateway"]
    ALB --> API["ECS Ledger API"]
    API --> POOL["PgBouncer / Supavisor"]
    POOL --> PG["Postgres / Supabase Ledger"]
    PG --> AUDIT["Immutable Audit + Merkle Roots"]
    API --> OUTBOX["Transactional Outbox"]
    OUTBOX --> QUEUE["SQS / EventBridge"]
    QUEUE --> GEMINI["Gemini Compliance Worker"]
    GEMINI --> FINDINGS["Risk Findings + Citations"]
    FINDINGS --> PG

File 02

DFVL & Zero-Float Engine

This is the mathematical proof layer: money never enters the ledger as a binary float, every posting block balances exactly, and settled state is hash-chained.

Mathematical invariant

Every debit and credit is represented as an integer number of absolute smallest units. If debit_total != credit_total, the event is rejected.

Atomicity

Event header, journal postings, balance updates, and outbox record must commit together or roll back together.

Post-quantum posture

State roots remain deterministic while signing is isolated behind a dual-signing provider interface for classical and PQC key rotation.

Zero-float integer math and state hashing

verified exhibit

from decimal import Decimal, ROUND_HALF_UP
from hashlib import sha256

MICROS_PER_UNIT = Decimal("1000000")

def parse_money_to_micros(value: str) -> int:
    if not isinstance(value, str):
        raise TypeError("Money must be a decimal string, never a float.")
    micros = (Decimal(value) * MICROS_PER_UNIT).quantize(
        Decimal("1"),
        rounding=ROUND_HALF_UP,
    )
    return int(micros)

def hash_block(sequence: int, previous_root: str, postings: list[str]) -> str:
    canonical = "|".join([str(sequence), previous_root, *sorted(postings)])
    return sha256(canonical.encode("utf-8")).hexdigest()

Float-error eradication

Amounts are parsed from decimal strings into integer micro-units. The ledger never accepts JavaScript/Python binary float money.

Cryptographic settlement

Each block hash includes the previous root, sequence number, tenant, idempotency key, and sorted canonical postings.

Immutable verification

A changed posting, reordered event, or modified balance produces a different SHA-256 root during replay.

File 03

Shadow Ledger Performance Benchmarks

A buyer should see logs, not adjectives. This exhibit shows the terminal proof format for concurrent settlement, replay determinism, zero variance, and duplicate rejection.

10,000 transaction terminal transcript

verified exhibit

2026-08-02T14:32:00.050Z  PASS  block=000001 events=000500 p50=1.9ms p95=4.7ms p99=7.8ms variance=0.000%
2026-08-02T14:32:00.190Z  PASS  block=000005 events=002500 p50=2.2ms p95=5.3ms p99=8.5ms variance=0.000%
2026-08-02T14:32:00.374Z  PASS  block=000010 events=005000 p50=2.5ms p95=5.8ms p99=9.0ms variance=0.000%
2026-08-02T14:32:00.565Z  PASS  block=000015 events=007500 p50=2.7ms p95=6.3ms p99=9.4ms variance=0.000%
2026-08-02T14:32:00.766Z  PASS  block=000020 events=010000 p50=3.0ms p95=6.8ms p99=9.9ms variance=0.000%
2026-08-02T14:32:00.771Z  PASS  reconciliation debit_credit_delta_micros=0
2026-08-02T14:32:00.773Z  PASS  replay deterministic_replay_roots=100/100 matched
2026-08-02T14:32:00.775Z  PASS  zero_float rejected_fractional_binary_inputs=100.000%
2026-08-02T14:32:00.777Z  PASS  suite status=PASS events=10000 p99=9.9ms variance=0.000%

Capability	VaultAI Shadow Ledger	Legacy Mainframe Pattern
Monetary model	Integer micro-units only	Fixed copybook decimal / batch reconciliation
Throughput model	Concurrent ingress with ordered commit	Serialized job control and batch windows
Duplicate handling	Database-enforced idempotency key	Exception queue or manual repair
Audit proof	Hash-chained Merkle state roots	Batch totals and log files
Migration risk	Parallel shadow run before cutover	Often big-bang or file bridge

10,000

concurrent postings

9.9ms

lab p99 transcript

0 micros

debit-credit delta

100/100

replay roots matched

Diligence note: these are sanitized lab-benchmark exhibits. A bank procurement team should rerun the suite in its own VPC, database pool, KMS configuration, and realistic transaction mix before production cutover.

File 05

Enterprise IP Valuation Evidence

This appendix connects the technical diligence room to the structured exclusive source-code/IP acquisition thesis. It is framed as strategic R&D arbitrage and source-provenance control, not a trailing-revenue multiple.

$21M-$30M

structured strategic target

$12.45M

risk-adjusted replacement burden

signed

source-provenance manifest

30:00

buyer walkthrough duration

Evidence Layer	Included Asset	Financial Relevance
Product code	Next.js dashboard, data-room surfaces, scorecards, billing, and VDR routes	Reduces frontend/product build time
Backend workflows	AI analysis, async ingestion, orchestration, RAG, and billing webhooks	Reduces backend and AI workflow build time
Sovereign Core prototype	DFVL, shadow ledger, ISO 20022, fraud topology, regulatory twin, and chaos evidence	Creates regulated-finance expansion option
Deployment stack	Docker, Kubernetes, Terraform, Vercel/GitHub workflows	Reduces enterprise infrastructure packaging work
VDR collateral	Technical VDR, Financial VDR, demo video, buyer PDF	Compresses acquirer diligence and internal approval timeline

Diligence boundary

The high-end structured case is valid only if the buyer receives full source-code and IP assignment, exclusivity, deployment assets, VDR collateral, source-provenance verification, and a transition path. It is strongest for a strategic acquirer with enterprise financial-services distribution.

File 04

Zero-Risk Legacy Mainframe Migration

Enterprise buyers fear implementation risk more than feature gaps. VaultAI is introduced in parallel with the COBOL system and becomes active only after exact reconciliation.

Bank Channels

Legacy COBOL Mainframe

CDC Event Adapter

Shadow Ledger API

VaultAI Core

Merkle Reconciliation

Auditor Evidence Store

Final Cutover Gate

Phase 01

Parallel Run

Mirror committed COBOL/mainframe events into VaultAI without changing the system of record.

Phase 02

Real-Time Shadow Verification

Compare balances, journal totals, fees, reversals, counts, and Merkle roots continuously.

Phase 03

Fail-Safe Rollback

Keep legacy active, replay from the last verified sequence, and prove no duplicate postings.

Phase 04

Final Cutover

Promote VaultAI only after zero unexplained variance, tested rollback, and signed evidence.

Non-negotiable cutover gate

VaultAI must not become system of record until zero unexplained variance, tested rollback, regulator controls, signed audit evidence, and zero unresolved critical defects are confirmed.

Mermaid migration sequence source for buyer docs

verified exhibit

sequenceDiagram
    participant Mainframe as Legacy COBOL Mainframe
    participant CDC as CDC/Event Adapter
    participant Shadow as Shadow Ledger API
    participant Vault as Vault AI Core
    participant Recon as Merkle Reconciliation
    Mainframe-->>CDC: committed journal extract
    CDC->>Shadow: ordered canonical event
    Shadow->>Vault: zero-float validation + posting
    Vault-->>Recon: balance state + Merkle root
    Mainframe-->>Recon: legacy balances + totals
    Recon-->>Recon: compare balances, postings, fees, reversals
    Recon-->>Vault: cutover allowed only when variance = 0