Portfolio dashboard
Applications in queue, disbursed this month, at-risk repayments - with aging buckets for anything pending past its window.
~/openarch/work/ledgerline-// case study
Ledgerline
download_pdf - Ledgerline case studyAn SME loan underwriting and lifecycle platform. Automated credit scoring, a single underwriter queue, and end-to-end loan tracking - replacing spreadsheet-based decisioning with an auditable runtime.
openarch.in/work/ledgerline
//the brief
What we built.
Northbridge underwrote on spreadsheets and email threads. We built one system that pulls bureau and bank-statement signals, scores every application into a risk band, and carries the loan from intake through disbursal to closure - with every status change written to an audit log.
domaincredit underwriting
shapeinternal platform
statusin production
surface7 core screens
stack
next.jsnode · expresspostgrespython · scoringredis
//what we shipped
Application intake
Applicant details, requested amount and purpose, plus document capture for PAN, GSTIN, bank statements and ITR.
Underwriting queue
Every application sorted by score band - Low Risk, Review, High Risk - with bulk assignment and document requests.
Credit score detail
Score broken down by factor: repayment history, balance stability, sector risk, bureau score. Bureau and statement viewers inline.
Loan lifecycle
A timeline from applied to closed, with the EMI schedule and paid / due / overdue states rendered per instalment.
Compliance & audit
Every status change and document access logged against an actor and a timestamp. Accountable by construction.
//data model
table applications
applicant_id, requested_amount, purpose, status, submitted_at
table bureau_reports
application_id, bureau_name, score, report_json, pulled_at
table credit_scores
application_id, model_version, score, band, factors_json
table loans
application_id, principal, tenor_months, rate, status
table repayments
loan_id, due_date, amount_due, amount_paid, status
// loan status machine
- 01applied
- 02underwriting
- 03approved
- 04disbursed
- 05active
- 06closed
//business logic
function scoreBand(score) { if (score >= 720) return 'LOW_RISK' // auto-approve eligible if (score >= 600) return 'REVIEW' // manual underwriter review return 'HIGH_RISK' // decline unless override}const weights = { bureauScore: 0.35, // external credit history statementStability: 0.30, // avg balance + bounce count yearsInBusiness: 0.15, sectorRisk: 0.10, repeatBorrower: 0.10,}//on the roadmap
Honest edges.
- →OCR fallback for scanned (non-native) bureau report PDFs
- →Partial-payment handling in bulk EMI reconciliation
- →Quarterly refresh of sector risk ratings from an external feed
//result
4d → 6h
underwriting turnaround, Low Risk
1
shared review queue, not email
100%
status changes audited
Northbridge cut average underwriting time from four days to under six hours for Low Risk applications, with every Review-band case routed to a single underwriter queue instead of scattered email threads.