Disclaimer: Independent educational project. Not affiliated with J.P. Morgan Payments or Mphasis. Built by Kaushal Khodifad. Data from public sources; figures are estimates.return to portfolio
Disclaimer: Independent educational project. Not affiliated with J.P. Morgan Payments or Mphasis.
Settlement & Reconciliation Engine - An independent, synthetic reference implementation of merchant-acquiring settlement & reconciliation. Modeled on J.P. Morgan Payments' position as the #1 US acquirer; not affiliated with, endorsed by, or representative of any system operated by J.P. Morgan / JPMorgan Chase or Mphasis. All figures are illustrative.
MerchantPay
Settlement & Reconciliation Engine · modeled on J.P. Morgan Payments
Modeled on J.P. Morgan Payments · Merchant-acquiring settlement & reconciliation infrastructure

J.P. Morgan Payments - Settlement & Reconciliation Engine

A glass-box engine where a synthetic multi-currency transaction book flows through a real fee waterfall, T+n settlement batching, and three-way reconciliation - and every headline number traces back to the rule and the row that produced it.

the #1 US merchant acquirer by purchase volume (TSG 2025 Directory)BFS delivery archetype: Mphasisseed 20260601
✓ ENGINE SELF-CHECK PASSINGInvariants 6/6 holdBlind adversarial challenge 8/8 above-tolerance caught · 0 FP
Money movement1Authorize
Effective take rate
2.50%
Σ(fees + tax) / Σ captured gross
click to audit ↗
Gross captured
$1.51M
5,532 captures · USD base
Net settled
$1.39M
80 payouts across 5 ccy
Recon match
96.84%
value-weighted, three-way
Open exception value
$43.9K
18 open breaks
FX spread revenue
$2.5K
spread on cross-currency settle
Net settled by currency
$1.39M · 5 settlement currencies
EUR$435.3K · 31.3%
GBP$347.0K · 25.0%
USD$305.2K · 22.0%
INR$284.8K · 20.5%
AUD$17.9K · 1.3%
Payout pipeline by value-date
T+1 / T+2 batches, business-day adjusted
06-02
$52.6K
06-03
$104.3K
06-04
$125.8K
06-05
$91.2K
06-08
$149.6K
06-09
$229.3K
06-10
$106.1K
06-11
$101.5K
06-12
$82.7K
06-15
$167.4K
06-16
$179.6K
T+1 T+2
Recompute
Perturb an input · watch every KPI move
re-ran in 0ms

Each control re-runs the full pipeline - generate → fees → settlement → reconciliation - on a fresh memoized pass. Constants can't do this; a real engine can.

Interchange bps shift+0 bps
FX spread50 bps
Reserve bps shift+0 bps
Bad-file multiplier
Settlement cycle
Seed (regenerate the book)
One-click scenarios - before → after
Buyer & ROI - ops economics at your scale
At $20B settled/yr: $12.64M recovered vs $103.6K of ops cost

The ratios (exception bps of settled value, breaks per $1B, auto-clear share) are measured live on this seeded book; everything else is your assumption - move it. The sandbox’s injected break rate is deliberately far ABOVE a healthy production book, so the incidence slider (default 5% of the sandbox rate) dials it toward reality. “Recovered” means surfaced-and-clawed-back at the recovery rate you set; we do not claim auto-cure. Synthetic extrapolation

Annual settled volume (assumption)$20B
Break incidence vs this sandbox (assumption)5%
Cost per manual break touch (assumption)$8
Recovery rate on surfaced value (assumption)40%
Exception value surfaced
$31.60M/yr
316.0 bps measured on the sandbox book × your 5% incidence assumption
Est. recovered
$12.64M/yr
at your 40% recovery assumption
Ops cost of working breaks
$103.6K/yr
12,948 manual breaks/yr · 14% auto-cleared before a human sees them
Net value
$12.54M/yr
recovered − ops cost, before license
Blind-band ceiling (the honest number)
≤ $2.00M/yr
the 1 bps tolerance IS the max undetectable leak rate - a policy property, not a sample. The adversary's below-tolerance plants ($1.82 this run) prove the band is real.

Defaults: interchange shift 0, FX spread 50 bps, reserve as-configured per merchant, clean settlement files (1× injection), seed 20260601. Effective take rate shown as percentage points (pp).

What is real vs simulated
No production settlement file has ever touched this system.
Real5 code paths
fee waterfall · settlement netting · three-way matcher · blind adversarial challenge · AI analyst (live OpenRouter, else labeled deterministic fallback)
Public-sourced5 external facts
each sourced fact on the Methodology tab carries a primary link
Synthetic1 seeded book
every transaction, merchant, rate table and $ figure - reconciled against files this demo built and perturbed itself

↑ Independent educational project by Kaushal Khodifad. J.P. Morgan Payments is a real company in the merchant-acquiring settlement & reconciliation (the #1 US acquirer) space; this design and the underlying prototype were built by Kaushal as a portfolio study. Not affiliated with, endorsed by, or representative of J.P. Morgan Payments's actual product. Data is from public sources. Figures are estimates.

Return to portfolioOpen the live demo