Zentis AuditOS-7B
Proprietary model — Zentis AI / Techvantage.ai
A domain-adapted version of Qwen2.5-7B-Instruct fine-tuned for internal audit tasks in BFSI and platform/e-commerce contexts.
Capabilities
- Audit finding generation with structured JSON output
- Regulatory clause mapping (FSSAI, CGST, DPDP Act, Consumer Protection Act)
- Anomaly classification from tabular audit data
- Multi-agent audit persona simulation (Lead Auditor, Skeptical Watchdog, Detail Perfectionist, Compliance Guardian)
- Control extraction from regulatory documents
Output Schema
All findings are returned as JSON arrays conforming to the AuditOS 2.0 finding schema:
- control_id, control_description
- evidence_summary
- regulator_clause_reference, policy_clause_reference
- finding_description
- evidence_sufficiency_score (0–1)
- regulatory_exposure_score (0–1)
- suggested_severity (High/Medium/Low/Observation)
- confidence_score (0–1)
- anomalous_items (list of specific data-grounded items)
Intended Use
Internal use within the Zentis AuditOS platform. Not for external distribution.
- Downloads last month
- 206