zLend: A Dual-Scope Cash-Flow Reconstruction Framework for On-Chain Credit Underwriting

Girish G N; Ashutosh Sahoo; Akshay SP; Gurukiran S; Dhanashekar Kandaswamy

q-fin.RMAug 17, 2026

zLend: A Dual-Scope Cash-Flow Reconstruction Framework for On-Chain Credit Underwriting

Authors: Girish G N, Ashutosh Sahoo, Akshay SP, Gurukiran S, Dhanashekar Kandaswamy

Organizations: Zeru AI

Abstract

Decentralized lending lacks a credit bureau: a borrower's capacity to repay must be inferred entirely from public on-chain activity, without income verification or a liability record. This paper presents zLend, a deployed cash-flow underwriting framework that reconstructs a wallet's daily balance history from raw token transfers and derives short-duration repayment-capacity signals from it. The reconstruction is performed twice per wallet, once restricted to a fixed stablecoin basket and once over all fungible transfers, on the premise that a wallet's total token holdings and its liquid, spendable balance are distinct quantities whose conflation misprices risk. From each series we derive liquidity coverage against a fixed loan size, cash-flow volatility and regularity, a drawdown-and-recovery statistic adapted from quantitative finance, and a recurring-counterparty detector that identifies salary-like payment cadence from transfer timing alone. The two views are then compared: a wallet with large aggregate holdings whose stablecoin reserve rarely covers the loan size is flagged as a liquidity mismatch irrespective of total wealth. We specify the pipeline formally, document the golden-master methodology used to verify a cross-language production migration to numerical tolerance 1e-9, and characterize the tier function's parameter sensitivity with an independent reimplementation validated to exact agreement (78 of 78 field assertions) against the deployed system's reference fixtures. Tier assignment is governed predominantly by the reference loan size, with four of six reference wallets changing tier across loan sizes from USD 10 to USD 25,000; the drawdown and coverage criteria bind on disjoint wallets, so neither subsumes the other; and no criterion in the tier rule is inert. zLend is deployed in production, informing real lending decisions via third-party API integrations.

Figures & tables

Panel of Figure 1 — Figure 1: Reconstruction pipeline, executed independently and identically for the stablecoin and total-wealth scopes.

Parameter	Role	Value
$L$	Reference loan size for coverage	$100
$\varepsilon$	Zero-balance threshold	$1
$f_{\mathrm{abs}}$	Absolute floor, outflow-concentration gate	$5
$f_{\mathrm{pct}}$	Median-relative floor, same gate	$0.05$
$k_{\mathrm{rec}}$	Distinct days for counterparty recurrence	$3$
$\tau$	Flat-band tolerance, trend classification	$0.01$

Table 1: Pipeline parameters and deployed values.

Severity	Liquidity mismatch	Flow mismatch
None	flag not raised	flag not raised
Low	$r<4$ and $\phi_{\mathrm{sc}}^{(30)}\geq 0.35$	$\rho\geq 0.2$
Medium	$r\geq 4$ or $\phi_{\mathrm{sc}}^{(30)}<0.35$	$\rho<0.2$
High	$r\geq 6$ or $\phi_{\mathrm{sc}}^{(30)}<0.2$	$\rho<0.1$

Table 2: Mismatch severity ladders, resolved most-severe-first: a wallet satisfying the High condition is High regardless of whether it also satisfies Low or Medium.

r=\widetilde{B}_{\mathrm{tw}}^{(30)}/\widetilde{B}_{\mathrm{sc}}^{(30)}

is the balance ratio;

\rho

is the stablecoin inflow share.

Signal	Tier	$\phi^{(30)}$	Zero events (90)	$\mathrm{MDD}^{(90)}$	Trend (30)
Primary ( $\mathrm{sc}$ )	Insufficient	$\|\mathcal{D}\|<30$
	Strong	$\geq 0.80$	$=0$	$\leq 0.35$	$\neq$ decreasing
	Moderate	$\geq 0.50$	$\leq 1$	$\leq 0.65$	—
	Weak	otherwise
Context ( $\mathrm{tw}$ )	Supportive	$\geq 0.80$	—	$\leq 0.50$	$\neq$ decreasing
	Cautionary	$\phi<0.50$ or zero events $>0$ or $\mathrm{MDD}>0.75$ (disjunctive)

Table 3: Tier definitions. Conditions within a row are conjunctive except where noted; rows are evaluated top to bottom and resolved on first match.

Panel of Figure 2 — Figure 2: Decision flow. The two scopes are scored independently into parallel signals; mismatch flags are computed from cross-scope ratios and qualify, but never override, the stablecoin-derived primary tier.

Wallet regime	$\|\mathcal{D}\|$	$\phi_{\mathrm{sc}}^{(30)}$	$\mathrm{MDD}_{\mathrm{sc}}^{(90)}$	Zero events	Primary tier
Steady, low volume	86	$1.00$	$0.020$	$0$	Strong
Recurring income	95	$1.00$	$0.092$	$0$	Strong
Short history	36	$1.00$	$0.250$	$0$	Strong
Severe drawdown, recovered	76	$1.00$	$0.950$	$0$	Weak
Repeated depletion	79	$0.40$	$1.000$	$5$	Weak
Thin stablecoin reserve	80	$0.00$	$0.250$	$0$	Weak

Table 4: Reference wallets, grouped by primary tier. All values are stablecoin-scope.

Panel of Figure 3 — Figure 3: Reconstructed stablecoin balance for the drawdown wallet. Coverage is $100\%$ across the trailing 30 days despite a 90-day maximum drawdown of $0.95$ .

Panel of Figure 4 — Figure 4: Both reconstructions for one wallet over the same 80-day history. The views diverge by roughly four orders of magnitude throughout.

Wallet regime	$10	$50	$100	$250	$500	$1k	$2.5k	$5k	$10k	$25k
Recurring income	S	S	S	S	S	S	S	S	S	W
Short history	S	S	S	S	S	S	M	W	W	W
Severe drawdown	W	W	W	W	W	W	W	W	W	W
Repeated depletion	W	W	W	W	W	W	W	W	W	W
Thin stablecoin reserve	S	W	W	W	W	W	W	W	W	W
Steady, low volume	S	S	S	S	S	S	S	S	W	W

Table 5: Primary tier as a function of reference loan size

L

. S

=

strong, M

=

moderate, W

=

weak.

Wallet regime	$0.15$	$0.25$	$0.35$	$0.50$	$0.65$	$0.80$	$0.95$
Recurring income	S	S	S	S	S	S	S
Short history	M	S	S	S	S	S	S
Severe drawdown	W	W	W	W	W	W	S
Repeated depletion	W	W	W	W	W	W	W
Thin stablecoin reserve	W	W	W	W	W	W	W
Steady, low volume	S	S	S	S	S	S	S

Table 6: Primary tier as a function of the strong-tier drawdown bound (default

0.35

Wallet regime	$\phi\geq 0.8$	zero $=0$	$\mathrm{MDD}\leq 0.35$	trend	Tier
Recurring income	$\checkmark$	$\checkmark$	$\checkmark$	flat	Strong
Short history	$\checkmark$	$\checkmark$	$\checkmark$	flat	Strong
Steady, low volume	$\checkmark$	$\checkmark$	$\checkmark$	flat	Strong
Severe drawdown	$\checkmark$	$\checkmark$	$\times$	flat	Weak
Repeated depletion	$\times$	$\times$	$\times$	decreasing	Weak
Thin stablecoin reserve	$\times$	$\checkmark$	$\checkmark$	flat	Weak

Table 7: Strong-tier criterion satisfaction under deployed parameters.

\checkmark

=

satisfied,

\times

=

fails.

Panel of Figure 5 — Figure 5: Reference wallets in the (coverage, drawdown) plane. Circles denote strong-tier wallets, squares weak. The severe-drawdown and thin-stablecoin wallets fail on orthogonal criteria, so neither statistic subsumes the other.

Wallet regime	$0.20$	$0.35$	$0.50$	$0.65$	$0.80$
Recurring income	$\times$	$\times$	$\times$	$\times$	$\times$
Short history	$\times$	$\times$	$\times$	$\times$	$\times$
Severe drawdown	$\times$	$\times$	$\times$	$\times$	$\times$
Repeated depletion	$\times$	$\times$	High	High	High
Thin stablecoin reserve	High	High	High	High	High
Steady, low volume	$\times$	$\times$	$\times$	$\times$	$\times$

Table 8: Liquidity-mismatch flag under the coverage-threshold sweep (default

0.5

, Equation 9 ).

\times

= not flagged.

Wallet regime	$0.05$	$0.10$	$0.15$	$0.20$	$0.25$	$0.35$	$0.50$	$0.75$
Recurring income	$\times$	$\times$	$\times$	$\times$	$\times$	Low	Low	Low
Short history	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$
Severe drawdown	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$	Low
Repeated depletion	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$
Thin stablecoin reserve	High	High	High	High	High	High	High	High
Steady, low volume	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$	$\times$	Low

Table 9: Flow-mismatch flag under the share-threshold sweep (default

0.25

, Equation 10 ). Severity is resolved most-severe-first: a wallet satisfying the High condition of Table 2 is High regardless of whether it also satisfies Low or Medium.

Explore similar work

Sep 8, 2026

zScore-N: A Neural Network for On-Chain Wallet Reputation Scoring

Wallet reputation scores decide who receives an airdrop, who can borrow, and who enters an allowlist across decentralised finance. They almost always begin as hand-written formulas: compositions of clamped logarithmic, linear and square-root transforms over behavioural features, with every threshold and point award set by hand. Such a formula is readable and deterministic, but it is piecewise and non-differentiable, it cannot improve as data accumulates, and it cannot distinguish a feature that is genuinely zero from one its pipeline failed to capture. We present zScore-N, the neural network that replaced ours in production. The formula served as its teacher: calibrated against 5,208,952 wallets sampled across 2019-2024 and verified to reproduce production output to within 2.3e-13, it supplies unlimited labelled training data at zero label noise. The trained network reproduces it to 0.58 points RMSE on the 1000-point scale (R^2 = 0.99997), against 2.25 for gradient-boosted trees and 28.04 for linear regression on identical features and splits. Trained with missing-value masks against uncorrupted targets, it halves the error that incomplete data introduces: at 10% feature-level missingness the formula drifts 51.4 points from its own complete-data output with a systematic -12.5 point bias, while the network drifts 17.9. The network carries the score at production scale, across a population of millions of wallets spanning six orders of magnitude in size and activity.

Girish G N, Ashutosh Sahoo, Akshay SP +2

Zeru AI

Decentralized Finance

Ask PDF

Jun 6, 2026

TRUST-SCF: Transformer-based Risk Understanding and Scoring for Transactional Supply Chain Finance

Supply Chain Finance (SCF) and LendTech platforms need credit scoring systems that respond to evolving transaction behavior, repayment delays, and active exposure. We propose TRUST-SCF, a transformer-based framework for transaction-level risk prediction and dynamic credit scoring. Each user history is represented as a sequence of transaction tokens containing utilization, repayment delay and transaction position. The main contributions are: (1) a financially aligned attention bias that combines utilization similarity and recency, enabling the model to compare repayment behavior under comparable exposure conditions; (2) continuous repayment-delay prediction in a log-transformed target space, reducing the influence of extreme delays while improving sensitivity to short-delay behavior and (3) a label-efficient credit-scoring pipeline in which the final credit score is not trained using any explicit external credit-score label, but is instead derived from predicted delay, potential risk over simulated utilization, actual unpaid exposure, and nonlinear calibration. Experiments on real transaction data from more than 300,000 transactions show that TRUST-SCF improves delay prediction over sequential baselines and produces scores that are strongly associated with future repayment behavior. These results suggest that TRUST-SCF is a practical framework for adaptive credit scoring and transaction-level risk mitigation in SCF and LendTech environments.

Mohammadamin Davoodabadi, Amirabbas Shakeri

Department of Growth Barook Co.

Financial Services AI Risk Management Credit Scoring

Ask PDF

Sep 23, 2026

FedIncome: Federated Learning for Income Estimation in Digital Lending Under Data Sovereignty Constraints

Verified income is often unavailable in digital loan applications, forcing lenders to rely on reported income and potentially leading to over-lending, overly conservative offers, or rejection of creditworthy applicants. Cross-institutional data-sharing constraints make this problem especially difficult for smaller lenders with limited training data. We introduce FedIncome, a federated learning framework for income estimation that enables institutions to train a shared model without pooling raw borrower records. Using more than one million LendingClub loans partitioned into

50

state-level clients, we simulate a heterogeneous lending consortium. The best federated model achieves out-of-time

R^2=0.608

, compared with

0.619

for a pooled centralised benchmark. Small-sample clients obtain an average out-of-time

R^2

improvement of

3.8

percentage points relative to the pooled centralised benchmark, while the fitted client-level relationship places the empirical crossover at approximately

4,790

training observations in this setting. When pooling is infeasible and the relevant alternative is local-only training, federation improves out-of-time performance across all sample-size groups, with the largest gains for data-scarce clients. We also combine federated income estimates with state- and income-specific debt-to-income thresholds. In a retrospective decision analysis, replacing reported income with the federated estimate increases simulated approval rates with only modest changes in observed default rates. FedIncome supports collaborative learning under data-locality constraints with little aggregate loss relative to pooled training and larger gains relative to local-only estimation.

Sultan Amed, Tanmay Sen, Sayantan Banerjee

OM & QT Area, Indian Institute of Management Indore, M.P. 453556, India. · SQC & OR Unit, Indian Statistical Institute Kolkata, W.B. 700108, India

Credit Scoring Finance Federated Learning Heterogeneous Federated Learning

Ask PDF