eKYC + AML · built for the agentic era

Compliance
you can prove.

Saolix builds explainable identity verification and financial-crime screening for banks and fintechs. It runs on your own infrastructure, in your own jurisdiction. AI assists the analyst. Every decision stays deterministic, auditable, reproducible.

Talk to us How we think →

Built by operators who have run payments and financial-crime systems at scale across Africa and Asia.

AI assists · probabilistic

Document read & liveness0.96

Sanctions name match0.88

Behaviour & device signal0.91

determinism boundary

Rules decide · deterministic · auditable

VERDICTCLEAR

rule_setv3.2 · pinned

auditsha256:9f3c…a7e1 · replayable

The problem

A regulator cannot audit a guess.

Most AI for compliance is a black box. It returns a score, not a reason — and when an examiner asks why you cleared a customer, “a model said so” is not an answer. Now AI can forge a flawless document in seconds and act on its own. The old controls are quietly failing. Saolix is built on the opposite principle.

How we think

AI assists the decision.
It never becomes the decision.

We draw a hard line. Above it, AI does what it is good at — reading documents, scoring risk, surfacing matches: fast and probabilistic. Below it, the things that must never be guessed — the identity verdict, the sanctions decision, the authorisation — stay rule-driven and logged. Same input, same output, every time. A decision you can replay in front of a regulator, byte for byte.

The determinism boundary: probabilistic AI signals above the line resolve into a deterministic, hash-stamped CLEAR verdict with full audit trail below it — The determinism boundary — probabilistic models inform above the line; a deterministic, hash-stamped verdict decides below it.

What we build

Two engines. One discipline.

Agentic AI on top, deterministic rules underneath. On-prem, quantized LLMs re-rank and explain, RAG keeps every reason grounded, and GenAI drafts the report. The verdict stays rule-driven, auditable, and reproducible.

Eagle · AML & sanctions screening

Eagle

Financial-crime screening that returns verdicts, not guesses.

A deterministic match chain produces byte-identical, reproducible results, with a tamper-evident audit trail and a regulator-replay endpoint that reconstructs any decision exactly. On top sits an agentic AI triage copilot: an on-prem, quantized LLM that re-ranks the alert queue and drafts RAG-grounded, plain-language explanations, with GenAI case reporting. It advises; the rules decide. Runs air-gapped, on your own infrastructure.

Agentic triage copilot — on-prem LLM, advises, never decides
RAG-grounded reasons & GenAI case reports
Reproducible verdict — same input, same output
SHA-256 audit row · regulator-replay endpoint

Known2Me (K2M) · agentic eKYC & identity

Known2Me

Verify the human — not the image of a document.

Document checks alone are finished: AI forges a perfect passport in seconds. Known2Me runs agentic eKYC and LLM- and RAG-based profile scoring, real-time and batch, leaning on the signals AI still cannot fake — liveness, behaviour over time, device and network provenance, and claims tied to a real issuing authority rather than a scanned file. GenAI drafts the report.

Agentic eKYC & profile scoring — real-time and batch
LLM & RAG based, liveness & behaviour signals
Claims-based, issuer-anchored · GenAI reporting

How it works

One pipeline. A hard line through the middle.

On the left, AI reads, resolves and ranks — fast and probabilistic. On the right, the verdict, the decision and the audit trail are rule-driven and reproducible. The only thing that crosses the line is a recommendation; the decision is always made on the deterministic side.

See it in practice

What a verdict actually looks like.

Every screening leaves an explainable record. The AI's probabilistic candidates and plain-language reasons sit on one side; the deterministic verdict and pinned ruleset on the other; the whole thing sealed with a hash you can replay in front of a regulator, byte for byte.

Real-time screening

Screening that keeps pace with the payment.

Every onboarding and every transaction is screened the moment it happens — names, sanctions, PEP and adverse media — and returns a deterministic verdict in milliseconds. Real-time or batch, it is the same engine and the same audit trail. The fast, probabilistic work runs the queue; the decision stays rule-driven.

EAGLE · LIVE SCREENING ILLUSTRATIVE STREAM

1,240SCREENED / SEC

38msMEDIAN LATENCY

99.2%AUTO-CLEARED

TIMESUBJECTSCREENED AGAINSTVERDICT

Why Saolix

Built to survive diligence.

Regulators and acquirers ask the same questions. We build to the answer.

Explainable

Every verdict carries its reasons, in plain language an examiner can read.

Reproducible

Byte-identical results and an immutable audit row, replayable on demand.

Sovereign

Runs on your infrastructure, on-device and offline-capable. Your data never leaves.

Disciplined

Reject-on-disagree, honest below the confidence floor, self-monitoring for drift.

Where to start

Start with the advisory.

Most teams don’t need another tool yet — they need a clear-eyed plan. We help banks and fintechs design eKYC and AML for the agentic era: where to add AI, where to keep humans, and how to keep every decision defensible. Engagements begin as advisory and grow into build.

Book a working session

Advisory
→ Build

consulting-first, by design

eKYCidentity

AMLscreening

Governancedefensible by design

Advisory & consultation

Where we advise.

The product is one way in. The other is sitting alongside your team across the whole compliance picture — strategy, regulation, technology and financial crime — drawing on careers spent running these systems in production, not reading about them.

01 · Compliance & regulatory

Compliance strategy

AML and CFT programme design, KYC and eKYC frameworks, risk-based policy and controls, and examiner-readiness — built for how regulators actually test you in the agentic era.

02 · AML & financial crime

Financial-crime advisory

Sanctions and screening design, transaction-monitoring tuning, risk assessments, model validation and false-positive reduction — decisions that hold up under audit, not just on paper.

03 · Technology & architecture

Technology & AI

Where to apply AI and where not to, on-prem and sovereign deployment, build-versus-buy, integration, and the data and audit infrastructure that has to sit underneath it all.

04 · Business & operating model

Operating model

Target operating model, analyst workflow design, cost-to-serve, scaling compliance operations and vendor selection — the business side of actually getting it live and keeping it there.

Most engagements begin here, as a focused working session, and grow into build only where it earns its place. Book a working session →

Built by operators

People who have run this in production.

Saolix is built by a team that has spent its careers operating payments, mobile money and compliance infrastructure at scale across Africa and Asia, not a vendor selling from the outside. We have taken AML and financial-crime screening to production, and we build that discipline into the product: AI assists the analyst, deterministic rules decide.

100M+users operated at scale

$1B+daily transaction value

AMLshipped to production

Sovereignon-prem & air-gapped by design

Complianceyou can prove.