● engine ready

Compliance Command Center

Live AML posture across your institution. All computation runs on-device.

0Risk score
LOW RISK

Composite of open alert severity, screening match rate and KYC portfolio mix. Updates every run.

Transaction volume (24h windows)
Top risk jurisdictions
Latest alerts

AI Transaction Monitor

Streaming AML engine: 7 rule families + unsupervised anomaly scoring (robust MAD z-score) with explainable verdicts.

Data input

Direction: in/out. Channel: upi | card | netbanking | cash | wire | crypto. Country: ISO code.

Engine settings (live)
Thresholds apply on the next run. Every verdict carries its rule provenance.
Alert queue

Advanced Analytics

Forensic layer: Benford first-digit conformance, transaction network graph with ring & hub detection, fan-in smurfing clusters and jurisdiction exposure matrix. All computed on-device.

Benford’s Law conformance
Forensic test used in real AML reviews: fabricated or manipulated amounts skew the first-digit distribution away from the Benford curve.
Jurisdiction exposure matrix
Volume weighted by FATF-style country risk weight. Exposure score drives the R2 rule set.
Transaction network
Gold dashed = layering ring (counterparty is itself a customer in the book). Orange ring = hub receiving from ≥3 customers. Red edges = high-risk corridor. Node labels: left = customers, right = counterparties.
Fan-in smurfing clusters

TBML & VASP Intelligence

Trade-based money laundering detection (over/under-invoicing, abnormal value-per-shipment, round-trip trade clusters) plus FATF Travel Rule (R.16) completeness on VASP/crypto flows. All computed on-device.

VASP & FATF Travel Rule exposure
R.16: originator and beneficiary data must travel with every VASP transfer. Add two optional CSV columns (originator,beneficiary) when pasting transactions to carry the data through.

Continuous Monitoring (24/7 Nazar)

Keyless 24/7 posture: the edge worker polls only PUBLIC status pages (no API keys, no credentials), and your audit-log exports (AWS CloudTrail, GitHub, Okta, CSV/JSON) are parsed and diffed entirely on-device. No user data leaves your browser; only the public status checks touch the network.

never checked
Add any public status URL
Audit-log ingestion (on-device)

Paste an export as JSON or CSV: AWS CloudTrail (Records[] or an event array), GitHub audit log export, Okta system log export, or any CSV with time/user/action/outcome columns. Nothing is uploaded; snapshots live only in your browser (localStorage).

Monitoring alerts

No monitoring alerts yet.

PEP Intel & Adverse Media

Detailed profiles for PEPs, sanctions designees and fugitive economic offenders: photo, role, country, legal status and source lists. Photos are public images from Wikipedia; profiles are summarised from public records (UN/OFAC/ED/court reporting) and are reference material, not allegations beyond those records. Everything stays on-device.

Add PEP from web (auto-enrich)
Type a name. RegShield pulls the public Wikipedia profile, photo and summary, then runs an on-device case study (watchlist cross-check + adverse media + EDD checklist). Nothing you type leaves the device except the public name lookup.
Adverse media scan (seeded demo corpus)

The fourth pillar of AML screening: negative-news matching. This demo runs against a seeded corpus of publicly reported items tied to the profiles above (fraud-seed entries are fictional demo items); the severity tag is the strongest claim in the item. Nothing is uploaded.

Sanctions & PEP Screening

Fuzzy name matching (normalized Levenshtein + token overlap + phonetics) against a REAL sanctions corpus: full OFAC SDN (19,456 entries, fetched 20 Oct 2026), MHA UAPA Schedule 1 organisations and Indian PMLA/fugitive-offender records. Non-Latin names (Devanagari and Urdu script) are transliterated to Latin automatically before matching (v3.6). All data on-device; no egress.

Screen customers
Watchlist corpus
NameTypeListCountry
Screening results
CustomerMatchScoreList / typeVerdict
Run a screening to populate.
Batch screening (50+)
#SubjectBest matchScoreList / typeVerdict
Run a batch to populate.
Ranked by match score. Verdicts use the match threshold slider: HIT at or above it, REVIEW 12 points below, CLEAR under that. CSV export includes every row.

KYC Risk Rating Engine

Weighted factor model (geography, product, channel, occupation, activity pattern). Editable weights recalibrate the whole portfolio instantly.

Model weights
Portfolio mix
Customer book
CustomerCountryOccupationScoreBandEDD
VASP & FATF Travel Rule exposure
Risk-scored VASP exposure per customer: crypto/VASP volume plus missing R.16 originator-beneficiary data.

Rule Engine

Every AML rule family is auditable: toggle, tune thresholds, and fire a dry-run against your current dataset.

RuleLogicActiveAlerts
Back-testing lab (30-day replay)

Replays the last 30 days of loaded transactions through the real engine, re-running it once per rule with that rule disabled, to give each rule's alert uplift and an estimated false-positive share from the 90-day expected-activity baseline.

Case Management

Escalate alerts into investigations with notes, priority and lifecycle tracking. Feeds SAR generation.

Regulatory Reporting

India-regime reporting desk: FIU-IND STR/CTR workflow, PMLA section references, SAR narrative drafting and exports.

CTR candidates (≥ ₹10 lakh single or structured below)
RefDateAmountCustomerChannel
Statutory calendar
SAR / STR narrative generator

Audit Trail

Every scan, verdict, override and export is logged with timestamp. PMLA/ISO 27001-friendly evidence trail.

TimeActorActionDetail

Settings & Data

Workspace controls. Everything is stored locally on this device; nothing leaves it.

Data portability
Deployment

RegShield AI v3.7 · real OFAC SDN screening (19,456 entries) · OFAC live-sync, batch screening, Devanagari/Urdu transliteration, compound-name matching (v3.6-3.7) · continuous monitoring (v3.1) · PEP Intel + adverse media (v3.2) · web-enriched PEP add + case study (v3.3) · edge-hosted PWA · offline-capable service worker · all risk logic on-device (no data egress).

Custom sanctions list (paste CSV)
Matching uses the same Levenshtein + token + phonetic engine as the OFAC/India lists and feeds rlBest directly, so custom entries screen at full strength.
OFAC SDN live sync
Never synced. The embedded corpus has 19,456 entries (fetched 20 Oct 2026).
Pulls the official public OFAC SDN CSV through the worker, diffs it against the last snapshot stored on this device, and lists new entries. Only the public CSV is fetched; nothing about your data leaves the device.
Fraud typology seed library
Preloaded synthetic fraudster seed list (240 entries) labeled by typology; used as an additional screening corpus alongside OFAC SDN, India curated and your custom list.