ORIGINATION
Loan approval
Eligibility gates, scorecard, risk-based pricing and a priced offer or a coded rejection.
Model any lending decision.
Ship it as one API.
Graviton BRE is a decision engine for lenders. Your credit team draws the policy as a visual graph of decision tables, simulates it on real files with a full trace, and ships it like code — versioned, approved and replayable, live on every channel in under a second.
No-code decision tablesSimulate before you shipEvery decision replayable
DECISION GRAPH · CVL-RETAIL● LIVE
Request
→
DECISION TABLE
Eligibility
↳
→
DECISION TABLE
Scorecard → Pricing
Decision
await graviton.evaluate('CVL-Retail', {
cibil: 742, foir: 38, ltv: 71
}) → 0.29s
cibil: 742, foir: 38, ltv: 71
}) → 0.29s
THE EDITOR · TRY IT LIVE
Send a file through your policy.
APPLICANT
BRANCH
INPUT
Request
APP-40192
OUTPUT
Decision
SANCTION
OUTPUT
Decline
Risk pricing · decision table
BAND → RATE + DECISION
Band
FOIR
Base rate
Decision
A
any
12.9%
auto-sanction
B
any
14.5%
auto-sanction
C
≤ 50%
16.9%
auto-sanction
C
> 50%
16.9%
refer → L2 (v8)
Not evaluated for this applicant — the path stopped earlier.
SANCTION
Auto-sanction
Band B · 14.5% · ₹12,00,000 offered · no human touch
TRACE
end-to-end 0.29s
▶ evaluate CVL-Retail@main · Standard
· request inputs assembled 0.02ms
✓ eligibility 5 rules · PASS 0.06ms
✓ scorecard 742 → band B 0.05ms
✓ pricing band B → 14.5% 0.04ms
✓ decision AUTO-SANCTION 0.29s
OUTPUT · JSON
{
"decision": "auto_sanction",
"band": "B",
"rate": 14.5,
"amount": 1200000,
"reasons": []
}
APPLICANT
BRANCH
INPUT
Request
APP-40192
OUTPUT
Decision
SANCTION
OUTPUT
Decline
Risk pricing · decision table
BAND → RATE + DECISION
Band
FOIR
Base rate
Decision
A
any
12.9%
auto-sanction
B
any
14.5%
auto-sanction
C
≤ 50%
16.9%
auto-sanction
C
> 50%
16.9%
refer → L2 (v8)
Not evaluated for this applicant — the path stopped earlier.
SANCTION
Auto-sanction
Band A · 12.9% · ₹12,00,000 offered · no human touch
TRACE
end-to-end 0.21s
▶ evaluate CVL-Retail@main · Clean prime
· request inputs assembled 0.02ms
✓ eligibility 5 rules · PASS 0.06ms
✓ scorecard 780 → band A 0.05ms
✓ pricing band A → 12.9% 0.04ms
✓ decision AUTO-SANCTION 0.21s
OUTPUT · JSON
{
"decision": "auto_sanction",
"band": "A",
"rate": 12.9,
"amount": 1200000,
"reasons": []
}
APPLICANT
BRANCH
INPUT
Request
APP-40192
OUTPUT
Decision
SANCTION
OUTPUT
Decline
Risk pricing · decision table
BAND → RATE + DECISION
Band
FOIR
Base rate
Decision
A
any
12.9%
auto-sanction
B
any
14.5%
auto-sanction
C
≤ 50%
16.9%
auto-sanction
C
> 50%
16.9%
refer → L2 (v8)
Not evaluated for this applicant — the path stopped earlier.
SANCTION
Auto-sanction
Band C · 16.9% · ₹12,00,000 offered · no human touch
TRACE
end-to-end 0.33s
▶ evaluate CVL-Retail@main · Borderline
· request inputs assembled 0.02ms
✓ eligibility 5 rules · PASS 0.06ms
✓ scorecard 712 → band C 0.05ms
✓ pricing band C → 16.9% 0.04ms
✓ decision AUTO-SANCTION 0.33s
OUTPUT · JSON
{
"decision": "auto_sanction",
"band": "C",
"rate": 16.9,
"amount": 1200000,
"reasons": []
}
APPLICANT
BRANCH
INPUT
Request
APP-40192
OUTPUT
Decision
REFER
OUTPUT
Decline
Scorecard · decision table
BUREAU SCORE → BAND
Bureau score
Band
≥ 760
A
720 – 759
B
700 – 719
C
< 700
— · decline
NTC / no score
— · manual (L2)
Not evaluated for this applicant — the path stopped earlier.
REFER
Refer → L2 deviation
new-to-credit — manual assessment · routed with justification
TRACE
end-to-end 0.38s
▶ evaluate CVL-Retail@main · Thin file
· request inputs assembled 0.02ms
✓ eligibility 5 rules · PASS 0.06ms
• scorecard NTC · no bureau score 0.05ms
• decision REFER · L2 deviation 0.38s
OUTPUT · JSON
{
"decision": "refer",
"queue": "L2_deviation",
"band": null,
"reasons": ["new_to_credit"]
}
APPLICANT
BRANCH
INPUT
Request
APP-40192
OUTPUT
Decision
OUTPUT
Decline
DECLINE
Eligibility · decision table
FIRST MATCHING RULE WINS
CIBIL
Active NPA
FOIR
LTV
Inflow
Gate result
any
true
—
—
—
KNOCKOUT · active NPA
any
false
> 55%
—
—
KNOCKOUT · FOIR ceiling
any
false
≤ 55%
> 80%
—
KNOCKOUT · LTV ceiling
any
false
≤ 55%
≤ 80%
< ₹40k
KNOCKOUT · inflow (v8)
any
false
≤ 55%
≤ 80%
≥ ₹40k
PASS → scorecard
Not evaluated for this applicant — the path stopped earlier.
DECLINE
Decline
active delinquency on bureau
TRACE
end-to-end 0.18s
▶ evaluate CVL-Retail@main · Adverse
· request inputs assembled 0.02ms
✕ eligibility KNOCKOUT · active delinquency on bureau
✕ decision DECLINE 0.18s
OUTPUT · JSON
{
"decision": "decline",
"reasons": ["active_delinquency"],
"counter_offer": null
}
APPLICANT
BRANCH
INPUT
Request
APP-40192
OUTPUT
Decision
SANCTION
OUTPUT
Decline
Risk pricing · decision table
BAND → RATE + DECISION
Band
FOIR
Base rate
Decision
A
any
12.9%
auto-sanction
B
any
14.9%
auto-sanction
C
≤ 50%
16.9%
auto-sanction
C
> 50%
16.9%
refer → L2 (v8)
Not evaluated for this applicant — the path stopped earlier.
SANCTION
Auto-sanction
Band B · 14.9% · ₹12,00,000 offered · no human touch
TRACE
end-to-end 0.29s
▶ evaluate CVL-Retail@v8-draft · Standard
· request inputs assembled 0.02ms
✓ eligibility 5 rules · PASS 0.06ms
✓ scorecard 742 → band B 0.05ms
✓ pricing band B → 14.9% 0.04ms
✓ decision AUTO-SANCTION 0.29s
OUTPUT · JSON
{
"decision": "auto_sanction",
"band": "B",
"rate": 14.9,
"amount": 1200000,
"reasons": []
}
APPLICANT
BRANCH
INPUT
Request
APP-40192
OUTPUT
Decision
SANCTION
OUTPUT
Decline
Eligibility · decision table
FIRST MATCHING RULE WINS
CIBIL
Active NPA
FOIR
LTV
Inflow
Gate result
any
true
—
—
—
KNOCKOUT · active NPA
any
false
> 55%
—
—
KNOCKOUT · FOIR ceiling
any
false
≤ 55%
> 80%
—
KNOCKOUT · LTV ceiling
any
false
≤ 55%
≤ 80%
< ₹40k
KNOCKOUT · inflow (v8)
any
false
≤ 55%
≤ 80%
≥ ₹40k
PASS → scorecard
Not evaluated for this applicant — the path stopped earlier.
SANCTION
Auto-sanction
Band B · 14.5% · ₹12,00,000 offered · no human touch
TRACE
end-to-end 0.29s
▶ evaluate CVL-Retail@main · Standard
· request inputs assembled 0.02ms
✓ eligibility 5 rules · PASS 0.06ms
✓ scorecard 742 → band B 0.05ms
✓ pricing band B → 14.5% 0.04ms
✓ decision AUTO-SANCTION 0.29s
OUTPUT · JSON
{
"decision": "auto_sanction",
"band": "B",
"rate": 14.5,
"amount": 1200000,
"reasons": []
}
APPLICANT
BRANCH
INPUT
Request
APP-40192
OUTPUT
Decision
SANCTION
OUTPUT
Decline
Scorecard · decision table
BUREAU SCORE → BAND
Bureau score
Band
≥ 760
A
720 – 759
B
700 – 719
C
< 700
— · decline
NTC / no score
— · manual (L2)
Not evaluated for this applicant — the path stopped earlier.
SANCTION
Auto-sanction
Band B · 14.5% · ₹12,00,000 offered · no human touch
TRACE
end-to-end 0.29s
▶ evaluate CVL-Retail@main · Standard
· request inputs assembled 0.02ms
✓ eligibility 5 rules · PASS 0.06ms
✓ scorecard 742 → band B 0.05ms
✓ pricing band B → 14.5% 0.04ms
✓ decision AUTO-SANCTION 0.29s
OUTPUT · JSON
{
"decision": "auto_sanction",
"band": "B",
"rate": 14.5,
"amount": 1200000,
"reasons": []
}
A live model of the Graviton decision engine — the same graph, tables and trace your team works in. Illustrative policy; your rules and thresholds are your own.
6 wks→1 day
to change a live policy
A vendor release cycle becomes an afternoon your credit team owns.
72%
of files auto-decided
Clear-cut files sanctioned or declined without a queue.
<400ms
median decision latency
Same speed at the branch, in the app, over the API.
100%
of decisions replayable
Inputs, node path and version stored on every single one.
Ranges reported by lenders after moving policy onto Graviton BRE. Your results depend on product mix and how much of policy you automate.
POLICY RUNS ON GRAVITON AT 48+ BANKS & NBFCs
SK Finance
Baid Finserv
MS Fincap
Universal Fingrowth
SAFL
Khush Housing
SIMULATE
Test before you deploy. Every time.
A single file shows the trace. Thousands show the risk. Backtest a draft against your own history, then let a challenger prove itself on live traffic — before it ever becomes the policy.
BACKTEST v8 · 24,000 FILES
● RUNNING
loaded 24,000 historic files · Jan–Jun
ran v8 vs every file · 0 errors · 41s
approve 61.2% → 63.8% (+2.6pp)
exp.loss 1.9% → 1.9% (flat)
migrate 1,204 files band B → C
match vs credit committee · 94.1%
flag 3 files · v8 declines, committee approved
✓ backtest complete · safe to challenge ▋
See exactly which nodes fired, which rows matched and what would have changed — before a single live borrower is touched.
Champion – challenger · live10% TRAFFIC → V8
CHAMPION · v7 · 90%
Approve 61.0%Auto 72%Early DPD baseline
CHALLENGER · v8 · 10%
Approve 63.6% ▲Auto 74% ▲Early DPD flat
6,120 live decisions on v8 · lift holding 9 daysPROMOTE WHEN READY
DEPLOY
Business rules are code. Ship them like code.
Branches to isolate a change, a full commit history, approval flows and one-click rollback — then one engine that runs the same rules everywhere you lend.
VERSION CONTROL
v8 · promoted to main · LIVE
approved by CRO · 21:04 · every channel
v8 · draft branch
tighten band-C FOIR · +inflow gate · backtested
v7 · retained
one-click rollback · replayable forever
Maker → CheckerAccept · reject · commentFull audit trail
ONE ENGINE · EVERY CHANNEL
const d = await graviton.evaluate(
'CVL-Retail', applicant
);
'CVL-Retail', applicant
);
Inside LOS · LMS · Gravicollect
REST API · app & partners
Cloud or on-premise
Sub-second · SLA-backed
No "branch version" drifting from the "app version" — one rule set answers every channel identically, the same second it ships.
TEMPLATES
Start from a decision, not a blank canvas.
One engine models every decision across the lending lifecycle. Fork a template, wire in your data sources and tune the tables to your policy.RISK · AML
Fraud & AML screening
Crime-scan, watchlist and behavioural checks that flag or hold an application in-flow.
PORTFOLIO
Portfolio risk monitor
Continuously score the live book against limits and trigger the right risk action.
SERVICING
Credit-limit adjustment
Re-price and re-limit existing borrowers on behaviour, exposure and bureau refresh.
COLLECTIONS
Collections treatment
Bucket-wise treatment paths, officer allocation and settlement eligibility as rules.
WORKFLOW
Deviation routing
An authority matrix that sends each deviation to the right approver with justification.
GOVERN · REPLAY
A regulator can replay any decision.
Every executed decision seals its inputs, the node path it walked, the rows it matched and the version that was live. Ask "why was this approved?" five months later and the answer replays on the exact policy that ran that day.
Decision #D-99148 · CVL-Retail@v7APPROVED · AUTO
Inputs — CIBIL 742 · FOIR 38% · LTV 71%STORED
Node path — eligibility → scorecard → pricingREPLAYABLE
Version — v7 by Head of Credit · approved CROMAKER-CHECKER
RUN IT WITH CONFIDENCE
Built for teams that answer to a regulator.
The facts your CRO, CISO and compliance heads ask for before a single live decision runs on it.Explainable by design
Every decision stores its inputs, node path, matched rows and version — replayable on demand.
Deployment, your way
Cloud or on-premise, single-tenant where you need it, with data residency in India.
Maker-checker & access
Every rule change is drafted, approved and versioned; MFA and role-based access throughout.
Real-time at any volume
Sub-second decisions embedded in journeys or over API, SLA-backed in production.
DATA SOURCES RULES CAN READ
Credit bureaus
Bank-statement analysis
KYC / CKYC
Valuation
Dedupe & exposure
Crime-scan / AML
GST / ITR
REST API
FAQ
Graviton BRE, in practice.
The questions credit, risk and IT heads ask before moving a live policy.
A decision engine for lending. You model eligibility, scorecards, pricing and deviations as a visual decision graph made of decision tables, execute it in real time across origination, servicing and collections, and every rule is versioned and every decision replayable.
Your credit team, with no code. Every change lives on a draft branch, is simulated against your historic files, approved via maker-checker (Head of Credit to CRO) and deployed as a new version with a full diff. The previous version stays live for one-click rollback.
Yes. Simulate any input and see exactly which nodes fired, which decision-table rows matched and how long each step took. Backtest a draft against thousands of historic files, then run champion-challenger on a slice of live traffic before you promote.
Yes. The same rule sets are exposed over one API — embedded inside Graviton LOS, LMS and Gravicollect, or called from your app, website and partner channels — so every channel returns identical decisions, typically under 400ms.
Multi-format ingestion: bureau responses, bank-statement analysis, KYC and CKYC, valuation, GST and ITR, internal exposure and dedupe, crime-scan and AML, scorecards and any custom fields your forms capture — all normalized to the fields your decision graph reads.
An authority matrix by amount, product and risk routes each deviation to the right level with a required justification, and every rule change moves through an approval flow — accept, reject or comment — with a full audit trail.
Yes. Model your current scorecards and policy as decision tables, simulate them against your own history to confirm they reproduce past decisions, then improve from there — you keep your credit thinking and gain versioning, testing and replay.
See your own policy run as a graph.
Bring one page of your credit policy — we'll model it live as decision tables, simulate it on your sample files, and show it deciding over the API.
✓Response within one working day
✓Runs inside Graviton or standalone over the API
✓Backtesting and champion–challenger included
Book a demo
Thirty minutes with a product specialist, on data shaped like your book. We respond within one working day.
Request a demoOr email neeraj@kugelblitz.xyz