dqagent

Built in Türkiye · KVKK-native · On-prem LLM

Banking-grade data quality platform for financial services — your data never leaves your server.

17 autonomous agents powered by a local LLM. BDDK, KVKK, BCBS 239, IFRS 9 and SOX compliant reports. A 14-phase pipeline in 22 seconds, with a 7-year encrypted audit trail.

Schedule a Demo Architecture & Security
  • On-prem deployment
  • Hardware-bound license
  • Zero data exfiltration
  • Turkish-native
dqagent.local · /dashboard

Total tables

1,284

+18

Active anomalies

23

-4

Quality score

92.4%

+1.2%

Compliance (5/5)

94%

+3%

SLA breaches

2

-1

Estimated impact

₺ 184K

-12%

Recent anomalies

14 phases · 22s

MUSTERI.TCKNCRITICALTCKN_FORMAT_VIOLATION47
HESAP.BAKIYECRITICALNEGATIVE_BALANCE3
ISLEM.TUTARWARNMULTIVARIATE_OUTLIER218
KREDI.VADE_TARIHIWARNFUTURE_DATE_VIOLATION12
FATURA.EMAILWARNEMAIL_FORMAT_VIOLATION89

Compliance status

KVKK96%
BDDK94%
BCBS 23991%
IFRS 988%
SOX100%

What is DQ-Agent?

DQ-Agent is an on-premise data quality platform for banks, insurers and other regulated financial institutions. Seventeen autonomous agents, powered by a local LLM that runs on the customer's own server, profile SQL Server databases, detect anomalies and produce BDDK, KVKK, BCBS 239, IFRS 9, SOX and AML compliance reports, without any customer data leaving the premises.

Deployment
On-premise only; there is no cloud-hosted mode
LLM
Local Ollama models; no calls to external LLM providers
Databases
SQL Server 2014+ (PostgreSQL and MySQL experimental)
Pipeline
14 phases orchestrated across 17 agents
Audit trail
7-year encrypted audit log
License
Annual and hardware-bound; the vendor has no remote access to customer data

Three foundational promises

Architectural principles, not marketing claims.

Data sovereignty

The LLM runs on your own server. No data or PII ever leaves. Licenses are bound to MAC + hostname; copied installations refuse to start.

Compliance lives in code

BDDK, KVKK, BCBS 239, IFRS 9, SOX and AML/5549 aren't PDF templates — they're agent behavior. Each anomaly maps to the relevant regulatory article.

Autonomous pipeline

14 phases, 17 agents. Profiling → anomaly → lineage → governance → RCA → forecast → cost. 22 seconds with LLM off, 7 minutes with LLM on.

A 14-phase autonomous pipeline

Each phase produces a snapshot; the next phase reads it. A single phase failure never stops the pipeline.

Phase 1

Schema Discovery

SchemaDiscoveryAgent

Catalogs all tables, FKs and finance-table patterns from INFORMATION_SCHEMA + sys.foreign_keys.

Typical duration
~5s
Minimum tier
professional
Tier:professionalenterpriseFinance Bundle

22 anomaly types in a single pipeline

Z-score, IsolationForest, format validation, drift, CDC and cross-table — all under one orchestrator.

NOT NULL violation

CRITICAL

NULL in PK column

High null rate

WARN

>30% NULL

Empty column

WARN

100% NULL

TCKN format error

CRITICAL

11-digit + check digit

IBAN format error

CRITICAL

TR + 24 digits + mod-97

Email format error

WARN

Non-RFC 5322

Future date

WARN

Transaction date > today

Numeric outlier

WARN

Z-score > 3

Multivariate outlier

WARN

IsolationForest

High IQR outlier rate

INFO

>5% rows

Negative balance

CRITICAL

Customer balance < 0

FK orphan

CRITICAL

Transaction → missing customer

Duplicate rows

WARN

Same PK twice

Cross-table inconsistency

WARN

Different NULL pattern across tables

Stale data

WARN

freshness > 24h

Null drift

WARN

+10% last week

Distinct shift

INFO

New category appeared

Volume drop

CRITICAL

-50% rows

Volume spike

WARN

+200% rows

DELETE spike

WARN

Unexpected deletes

Finance table DELETE

CRITICAL

Any delete

Reconciliation breach

CRITICAL

GL ≠ subledger

7-day SLA forecast breach

WARN

Linear regression

Five regulations, one platform

PDF + HTML reports per framework, with evidence chains mapped to specific articles.

KVKK

Turkish Personal Data Protection Law (Law 6698)

PII inventory, retention policies, masking and anonymization evidence.

MaskingAgent + ComplianceAgent

BDDK

Turkish Banking Regulator — Data Governance

Governance structure, ownership matrix, audit trail, lineage documentation.

GovernanceAgent + LineageAgent + ComplianceAgent

BCBS 239

Basel Risk Data Aggregation Principles

Per-principle (P1–P11) agent-based evidence: accuracy, completeness, timeliness, adaptability.

ProfilingAgent + AnomalyDetectionAgent + LineageAgent

IFRS 9

Financial Instruments — Expected Credit Loss

PD/LGD/ECL data completeness, timeliness and outlier checks for loan tables.

ProfilingAgent + ContractAgent + ReconciliationAgent

SOX

Sarbanes-Oxley — Information Controls

Encrypted 7-year audit log, role-based access, approval-gated mutations — supports SOX 302/404.

OrchestratorAgent (audit) + auth_manager

AML / 5549

Anti-Money-Laundering (Turkish Law 5549)

KYC data completeness, anomaly detection on suspicious transaction tables.

AnomalyDetectionAgent + CrossTableAgent

All compliance details

Dashboard included. Zero build pipeline.

A single HTML file, no Node toolchain. 13 tabs: KPIs, anomalies, lineage, compliance, heatmap, approvals, AI copilot.

  • 6 KPI widgets
  • 13-tab analytics
  • ISO 8000 heatmap
  • WebSocket live alerts
  • PDF / HTML / Excel export
  • Natural-language DQ queries

ISO 8000 Heatmap

Completeness

96%

Accuracy

92%

Validity

89%

Consistency

94%

Timeliness

87%

Uniqueness

98%

30-minute live demo against your own schema

Not a slideshow — a live install. We profile your SQL Server connection in minutes.

Schedule a Demo