SENTRA
AI Credit Decisioning for Banks, NBFCs & Fintechs
SENTRA is credit-risk infrastructure for regulated banks, NBFCs and fintechs — a real-time, explainable decisioning engine that computes PD and LGD from 500+ behavioral and financial signals, classifies assets to IRAC/SMA, runs Basel and stress-testing, and logs an immutable, regulator-ready audit trail for every decision. Fully aligned to RBI's FREE-AI framework, ECL, IRAC and Basel III.
What SENTRA does
Purpose-built capabilities, engineered to the regulation and ready for production.
Explainable AI engine — SHAP for global importance and LIME for per-decision rationale, with auto-generated adverse-action reason codes
PD & LGD computation from 500+ behavioral and financial signals using gradient-boosted ensembles, with dynamic recovery forecasting
Real-time PD/DPD decisioning — automated approvals, risk-based pricing, tenure recommendation and DPD bucketing for collections
IRAC, SMA & Basel automation — asset classification, early-warning SMA flags, Provision Coverage Ratio and Capital Adequacy Ratio per RBI mandates
Integrated stress testing — macroeconomic shocks, sectoral downturns and interest-rate volatility modelled against repayment capacity
Regulator-grade audit trails — every input, transformation and model weight logged immutably, with exportable evidence packs and the seven FREE-AI Sutras
Inside SENTRA
The core engines and workflows that make up the platform — each one an auditable control, not a promise.
Dual-framework attribution — SHAP for global importance, LIME for per-decision rationale — with plain-language adverse-action reason codes and immutable, exportable evidence packs.
Gradient-boosted ensembles estimate Probability of Default and model Loss Given Default across 500+ behavioral and financial signals, with dynamic recovery forecasting.
Real-time approval automation on live PD thresholds, risk-based pricing and credit limits, tenure recommendation, and DPD bucketing that triggers collections actions.
Automated Standard / Sub-standard / Doubtful / Loss classification, early-warning SMA flags, Provision Coverage Ratio and Basel Capital Adequacy Ratio per RBI mandates.
Simulates macroeconomic shocks, sectoral downturns and interest-rate volatility to expose concentration risk and repayment-capacity sensitivity.
Population Stability Index tracking, drift detection and performance-review triggers under a three-lines-of-defence model with versioning and rollback.
Reference metrics
Reference figures from model validation; actuals vary by portfolio and configuration.
Opaque, un-auditable credit decisions and unmanaged bias risk.
What you get out of it
Full RBI FREE-AI alignment — board-approved AI policy, fairness audits and contestable, documented rationale for every decision
Reference model performance of up to 0.87 AUC, with 30–40% NPA-reduction potential versus legacy scorecards
Real-time decisions at scale — ~28ms median inference, 4,200+ requests/sec, under a 100ms SLA
Model governance across three lines of defence — versioned models, approval gates, rollback and PSI drift monitoring
Seamless core-banking integration with Finacle, TCS BaNCS, Temenos and Finastra via secure APIs
Private, secure deployment — on-prem or VPC, mTLS, RBAC/IAM and data-residency controls (SOC 2 ready)
See SENTRA in a live demo
We'll tailor the walkthrough to your sector, data and regulators.