SaolixGroup
Saolix Eagle. AML screening with a record you can prove.
Sanctions and PEP screening where rules make every decision, AI helps the analyst, and each interactive screening is sealed in a tamper-evident record. Deployed inside your own perimeter.
The simulator runs in your browser with fictional cases and curated public UN references; no sign-in needed. The sample report uses synthetic data.
Each interactive screening decision is sealed with its record, and the record shows if it was ever changed.
Compliance teams drown in false positives. An examiner does not want a probability. They want to know exactly why an alert fired, and to trust the record that says so. Eagle is built for that question first. A fixed, deterministic match chain decides. AI can help explain and label alerts, but can never close one. Each interactive screening is sealed in a tamper-evident audit log whose integrity can be checked at any time.
Every reason visible.
Eagle connects matching, investigation, rules and evidence in one workflow, giving compliance teams the context to act and the record to defend that action later.
No date of birth or ID supplied, so the verdict is capped at review.
Rules, not a model, produce the verdict.
Illustrative product view with sample data. Not a customer record; identifiers and version numbers are examples.
AI assists above the line. Rules decide below it.
The determinism boundary is the core of the design. Whatever the copilot suggests, the verdict comes from pinned rules, and for interactive screening the audit record is written in the same transaction as the decision.
- ScreenCheck a person or entity against the configured sanctions and PEP data.
- ResolveBring identity details and matching evidence together.
- InvestigateSurface the context, relationships and records an analyst needs.
- Decide and preserveApply the rules, record the outcome and seal it in the audit log.
What Eagle does.
Deterministic match chain
Each name is normalised and transliterated. Candidate matches are found with trigram similarity (and, where enabled, a character-level embedding), then scored with a deterministic lexical measure that combines edit distance and token overlap. A near-exact match, or close Jaro-Winkler similarity together with agreement in how the names sound (Double Metaphone, Daitch-Mokotoff), can raise a score to a set floor, never lower it. The result is mapped to a verdict by configured thresholds, which can only be changed with a second person's approval. The decision logic is deterministic: the same candidates, under the same rules and lists, always give the same verdict.
multi-stage · transliteration-awareExplainable matching
Each match brings together name similarity, how the names sound, identity details such as a date of birth, and links to listed parties, so an analyst can see why a name matched.
match · resolve · explainCase investigation
Review potential matches, work through the evidence and record the analysis. Cases resolve through an approvals queue: by default, cases with a match score of 80 or more need a second person to approve, and Eagle can be configured so that the second approval on those cases needs an MLRO or administrator. A regulatory filing workflow is included.
triage · review · resolve · fileTamper-evident audit
Each interactive screening is written to a tamper-evident audit log, built on SHA-256 and keyed hashes, in the same transaction as the decision. Batch runs are recorded as a single batch entry in the general audit log. An integrity check recomputes the log on demand and reports whether it still verifies.
interactive: screen → case → audit · one transactionGated AI agent
In the analyst screen, Eagle's triage copilot explains each alert deterministically, drawing on the network graph and on similar resolved cases (retrieval-augmented triage). Raia, an optional AI agent available through Eagle's API, uses a locally hosted large language model (LLM) to label an alert as likely noise or needing review, with a confidence and a reason. Neither can close an alert, and a real hit is always sent back for human review.
retrieval · optional LLM agent · a person decidesSanctions feeds & rescreening
Eagle ingests sanctions data from sources including the OFAC, UN, EU and UK lists, and PEP data and some further lists from OpenSanctions (whose licence restricts commercial use, so customers need their own licence for them). It can be configured to rescreen earlier names when a list changes. Each case can be documented in a Case Intelligence Report.
OFAC · UN · EU · UK · configurable rescreenWhat comes next.
Quantitative risk signals
Quantitative finance and econometric methods to add structured risk context to a case. Like every AI signal in Eagle, they will inform the analyst and never decide.
Runs inside your walls.
Eagle is delivered as software you operate, not a service you send data to.
Fits into the systems you already run.
Eagle connects to onboarding, payments and case systems through standard interfaces, and hands your compliance team the records they need.
Screening API
Screen a customer or payment party from your own systems through Eagle's REST API. Each response carries the verdict (pass, review or fail) and the reasons behind it.
REST · verdict with reasonsBatch screening
Screen a whole customer base or file in one run, for example after onboarding a portfolio. Each batch run is recorded in the general audit log as a single batch entry.
batch runs · audit entryRegulatory filing
The regulatory filing workflow supports report formats including goAML suspicious and cash transaction reports, the format used by many financial intelligence units, and FinCEN suspicious activity reports.
goAML STR / CTR · FinCEN SARAudit and oversight
Administrators can run integrity checks on the whole screening log, the general audit log or a single request, and export the general audit log as a CSV report for review.
integrity checks · CSV exportKnow exactly what you screened against.
Eagle loads the major sanctions lists from their publishers, keeps them current, and records which version of each list was used for every interactive screening. Your compliance team decides which lists apply to your institution.
Official sanctions lists
Loaded directly from each publisher: the UN Security Council Consolidated List, the US OFAC Specially Designated Nationals list, the EU financial sanctions list and the UK Sanctions List.
UN · OFAC · EU · UKPEPs and further lists
Politically exposed persons, the Australian and Canadian sanctions lists and published INTERPOL Red Notices, through OpenSanctions. Its licence restricts commercial use, so customers arrange their own licence for this data.
via OpenSanctions · licence requiredClean updates
Eagle checks whether each list has changed before downloading it. Each update is applied in a single database transaction that adds new entries, updates changed ones and removes delisted ones, and is stamped with a version.
one transaction · versionedFreshness on record
Every interactive screening records which list versions were in force and how old each list was at that moment. Re-screening of earlier names when a list changes can be switched on.
list version · list ageNames written in Arabic, Cyrillic, Chinese, Korean and Japanese scripts are transliterated before matching, so names can be compared across scripts. How matching works across scripts →
Don't take our word for it. Inspect it.
Everything on this page can be demonstrated. Start with any of these.
Eagle Simulator
Walk through prepared screening and investigation cases: results, case evidence, the decision and the audit trail. No sign-in needed.
Open the simulator ↗ Download · PDFSample case report
A complete Case Intelligence Report for a synthetic case: match scores, network exposure, AI triage notes and the audit trail an examiner would see.
Download the sample → Download · PDFSample audit chain
Four synthetic screenings sealed with Eagle's linear-chain method, with real hashes and code to recompute them yourself.
Download the chain → On requestLive walkthrough
Book a working session and we will screen names, open the case and run the audit integrity check while you watch.
Book a session ↗Further reading.
The determinism boundary in practice
How Eagle lets AI help an analyst without ever letting it decide.
AuditTamper-evident by construction
How Eagle writes, seals and checks every screening record.
Sanctions screeningSanctions screening across alphabets
How Eagle normalises, compares and scores a name.
Your next chapter.
Let’s build it together.
Explore a product, discuss a project, or find the right expertise. Tell us where you want to go. We’ll help you take the next step.