Every architectural decision in FINVIJ starts from the same constraint: the data never leaves the bank, and every decision must survive an examination. On-premise and private-VPC deployment — by design, not retrofitted.
Borrower data, bureau responses, models, and prompts are processed entirely inside your infrastructure. Nothing is sent to third-party cloud services — including the AI layer, which runs on local models within your perimeter.
Bare metal, private cloud, or your AWS/Azure/GCP tenant. Single-tenant in every deployment. The platform was architected for this from day one, not adapted to it.
All data, all processing, all model inference stays within Indian jurisdiction — aligned with RBI expectations and the DPDP Act 2023.
Consent-aware data handling, purpose limitation, and data-principal rights supported in the data layer — not bolted on through policy documents.
Every decision traces end-to-end: policy rule → data source → model version → override log. Any past decision can be reproduced bit-for-bit for any date — examination-ready by construction.
RBAC with maker-checker separation on every change — policies, models, rate cards, and overrides all carry dual control and an append-only change log.
Deployment, access, and audit models designed to satisfy RBI's directions on IT outsourcing and third-party risk — because the system runs inside your perimeter, the outsourcing surface is minimal.
ECL staging, provisioning, and asset-classification outputs mapped to Ind AS 109 and IRACP norms out of the box — board-grade and examiner-legible.
Encryption in flight (TLS) and at rest (AES-256). Immutable, append-only audit logs. Model documentation packs auto-built for MRM review. Aligned with RBI MRM guidelines, Ind AS 109 / IFRS 9, ISO 27001 and SOC 2 control frameworks.
We would rather answer the perimeter questions on day one than on day ninety.
Request Demo