Machine intelligence for finance
Building intelligence
for finance.
Machine learning infrastructure for modern financial systems.
ELVNFI builds the infrastructure financial institutions use to understand risk, behaviour and money movement. Our first product, Risk-NG, is open-source risk intelligence for fraud detection and AML.
Built for Nigeria. Designed for Africa.
Transaction risk
Transfer · txn_8f21 · beneficiary age 3h
Risk signals
- Velocity anomaly+28
- New beneficiary+21
- Device deviation+19
- Unusual amount+17
- Network exposure+9
Decision stream
LAST 60s- Rules fired
- 3
- Model version
- v0.4.2
- Latency
- 18ms
- Decision
- REVIEW
Illustrative interface. Sample data.
Our first product
Risk, with context.
Open-source risk intelligence for financial institutions.
Risk-NG helps financial institutions detect fraud, monitor AML risk and understand suspicious behaviour across transactions, accounts, devices and financial networks.
Risk isn’t a property of a transaction. It’s a property of behaviour.
Transaction monitoring
Score every transfer, payout and collection in line, before the money leaves.
AML & screening
Watchlist matching, structuring detection and typology coverage on the same engine.
Behavioural intelligence
Per-account baselines for amount, timing, velocity, device and beneficiary patterns.
Case management
Alerts, queues, analyst decisions and a full audit trail of every score.
The problem
Financial crime doesn’t look like a rule.
It looks like a pattern.
A transaction rarely tells the whole story. Risk emerges from the relationships between the entities around it.
Swipe to explore → · Illustrative graph. Sample entities.
The risk engine
One engine.
Every signal.
Input
Transactions
transfers · card · collections · payouts
Stage 01
Ingestion
stream + batch
Stage 02
Feature engineering
velocity · baselines · graph features
Deterministic
Rules
Probabilistic
ML models
Relational
Network risk
Unified
Risk engine
single score · full explanation
Rules × ML
Rules catch what you know.Machine learning finds what you don’t.
# rule: large_transfer_to_fresh_beneficiary
IF
transaction.amount > ₦5,000,000
AND
beneficiary.age < 24h
THEN
risk += 30
# fired on txn_8f21 → +30Written, reviewed and versioned by your risk team. Every fire is traceable to a line of code.
- Behaviour deviation87%
- Velocity anomaly91%
- Device deviation84%
- Network exposure79%
Learned from your own transaction history. Surfaces the shapes nobody wrote a rule for yet.
Deterministic when you need control. Probabilistic when patterns become complex.
Illustrative interface. Sample data.
Open source
Your risk infrastructure shouldn’t be a black box.
A risk decision you cannot explain is a risk decision you cannot defend — to your board, to your regulator, or to the customer whose transfer you just held. Risk-NG is open source.
Built for Nigeria
Built for Nigeria.Designed to scale across Africa.
Financial infrastructure in emerging markets has its own transaction patterns, payment rails and operational realities — instant transfers that settle in seconds, agent networks that move physical cash, wallets that behave nothing like accounts.
A risk model trained on card-present fraud in another market does not understand any of it. Risk-NG starts by understanding those realities. Then expands globally.
Money movement
Bank
core banking
Wallet
stored value
NIP
instant transfer
Merchant
acceptance
Agent
cash-in / cash-out
Bank
settlement
Developers
Built to be used, inspected and extended.
One endpoint to score a transaction. A signed decision, the signals behind it, and the model version that produced it — every time.
{
"transaction_id": "txn_8f21",
"amount": 2450000,
"currency": "NGN",
"beneficiary_age": "3h"
}{
"risk_score": 0.94,
"decision": "review",
"signals": [
"velocity_anomaly",
"new_beneficiary",
"device_network_risk"
],
"model_version": "v0.4.2"
}Illustrative API shape. Not a published contract.
Risk-NG waitlist
Be early.
Risk-NG is being built in the open. Join the waitlist to get early access, product updates and invitations to the first private releases.
No spam. Just product updates.