Sentra

Detects the real intent behind transactions for African financial compliance.

Type: AI & automation
Status: In development

Results

847 ms average end-to-end transaction analysis

How it's built

Clients
- Compliance dashboard
  React, TypeScript
- Partner integrations
  Signed webhooks
Gateway
- API gateway
  Go, JWT, role-based access, rate limiting
Services
- Risk engine
  Rust scoring and business rules
- AI service
  Python, Groq, Llama 3.2
Messaging
- Redis
  Message broker and cache
Data
- PostgreSQL
  Transactions and audit log

Simplified overview. Requests flow from top to bottom. Services run as containers under Docker Compose; webhooks are signed with HMAC-SHA256.

Overview

Rule-based compliance catches what it's told to look for. Sentra looks for intent: masked P2P crypto trading, structuring under CBN reporting thresholds (₦5M individual, ₦10M corporate), money mule sweeps, Ponzi schemes, romance scams and BEC attacks. It's built as microservices: a Go API gateway, a Rust risk engine, a Python AI service on Groq and Llama, and a React compliance dashboard.

What it does

Intent detection for structuring, sweep fraud, masked P2P crypto and shadow-economy activity
CBN reporting thresholds and account tier validation built in
Multi-currency: NGN, USD, GHS, KES, ZAR, XOF/XAF, GBP, EUR, AED
Rust scoring engine for fast risk evaluation
LLM-generated explanations through Groq and Llama 3.2
HMAC-SHA256 signed webhooks and per-key rate limiting
847 ms average end-to-end analysis

Built with

Go
Rust
Python
React
TypeScript
PostgreSQL
Redis
Docker
Groq