[SF-11] EWA LENDING

dashboard
app shinhan a.i chat
app shihan for employee
admin - risk management 1
admin - risk management 2
admin - check business (accountant)
admin - index dashboard
admin - check customer (employee)
admin - business health

Inspiration The traditional consumer lending and salary-advance market is fundamentally flawed. It relies on manual, paper-based underwriting, leading to high acquisition costs (CAC) and delayed processing times. Worse, it often pushes vulnerable workers into a "debt trap" through predatory interest rates. We envisioned a shift from this reactive lending model to proactive Embedded Finance.

We were inspired to build an "Impartial AI Advisor"—a system that creates a definitive Win-Win-Win ecosystem:

For Financial Institutions: Secures a stable, long-term user base with Non-Performing Loans (NPL) strictly under 2% through source-controlled cash flows.

For Employers (HR): Boosts employee retention and provides macro-level insights into the workforce's collective financial wellness.

For Employees: Offers instant, zero-friction access to their earned wages while providing empathetic, unbiased financial guidance to foster sustainable financial health.

What it does SF-11 is an AI-powered Earned Wage Access (EWA) and Salary-Linked Lending platform. It eliminates traditional income documentation by integrating directly with corporate HRM and Payroll systems via real-time APIs.

Beyond simply dispensing funds, SF-11 acts as a Financial Butler:

Smart Limits & Hyper-Personalization: Instead of offering a flat 30% withdrawal limit, the AI analyzes individual spending patterns and upcoming bills to recommend a safe withdrawal amount (e.g., "You should only withdraw $50 now to ensure you can cover next week's utility bill.").

Impartial Financial Advisory: It assigns a dynamic "Financial Wellness Score" and provides actionable, personalized tips to help users avoid the debt cycle.

Automated Cross-Selling: If the AI detects a user constantly maximizing their EWA limit, it proactively suggests converting the short-term advances into a structured, lower-interest Salary-Linked Loan to help them consolidate debt.

How we built it We built a robust, multi-tier architecture utilizing Alibaba Cloud for scale and the Qwen model as our core AI brain.

Cloud Infrastructure (Alibaba Cloud): We utilized Alibaba Cloud's API Gateway and Message Queue services to handle asynchronous, high-volume transaction spikes (e.g., during holidays or back-to-school seasons) without system downtime. The core reconciliation engine uses a Split Payment architecture routing through Escrow accounts to instantly deduct fees/principals before dispersing the remaining salary.

Data Ingestion: Implemented an Event-Driven Webhook architecture to capture real-time HRM signals (clock-ins, resignations, disciplinary actions) rather than relying on delayed batch processing.

The AI Brain (Qwen & ML): * Conversational Interface: We integrated the Qwen LLM to power the natural language interactions of the Financial Butler. Users can simply type, "I need $20 for groceries," and Qwen parses the intent, checks the accrued wage balance, and processes the command contextually.

Risk Prediction: We deployed machine learning models for Churn Prediction (identifying "ghosting" behavior before HR files paperwork by analyzing metadata) and Anomaly Detection to dismantle organized Fraud Rings (e.g., spotting multiple ghost employees sharing exact clock-in timestamps and IP addresses).

Challenges we ran into Deterministic Limits vs. Human Unpredictability: Initially, our hard-coded Rule Engine worked perfectly for legal compliance (capping EWA at 30% of monthly salary). However, deterministic algorithms are blind to sudden human behavior changes—like an employee planning to quit the next day ("ghosting"). We had to pivot to probabilistic AI models to dynamically adjust credit limits in real-time based on behavioral anomalies.

Liquidity Crunches: Ensuring real-time disbursement requires having enough idle capital in the Escrow account. Forecasting this need manually was inaccurate. We had to build an AI Cashflow Forecasting module to predict macro-level withdrawal spikes based on seasonality and sector-specific trends.

Balancing UX with Anti-Fraud: We needed the transaction to be frictionless ("money in seconds") but secure against Account Takeover (ATO) attacks. Implementing seamless Biometric Authentication tied to Device Binding was a complex but necessary hurdle.

Accomplishments that we're proud of We are incredibly proud of transforming a cold, transactional lending tool into an empathetic financial companion. By leveraging Qwen, we successfully bridged the gap between strict risk management and personalized user care.

Furthermore, designing the Split Payment/Virtual Account reconciliation flow was a major technical milestone. This mechanism intercepts the corporate payroll before it hits the user's personal bank account, automatically settling the EWA debt. This architecture practically eliminates the risk of default, allowing us to confidently project an NPL of <2%.

What we learned We learned that in fintech, automation solves the problems of the present, but AI solves the problems of the future. Relying solely on clean data and logical IF/THEN rules is insufficient when scaling. Systems suffer from "Garbage In, Garbage Out" if HR delays updating their databases. We realized that AI's true value isn't just in crunching numbers—it's in reading between the lines of unstructured data, recognizing behavioral patterns, and providing scalable empathy to end-users who need financial guidance, not just a quick loan.

What's next for SF-11 (Bonus) Corporate Dashboard Enhancements: Upgrading the employer portal with an AI-generated "Heatmap of Collective Financial Stress." This will allow HR departments to visualize financial anxiety across different branches or departments and intervene with targeted welfare programs.

Expanded Qwen Capabilities: Deepening the integration with Qwen to support multi-lingual conversational interactions, catering to diverse, blue-collar migrant workforces in large industrial zones.

Macro-Economic Integration: Feeding real-time global supply chain and economic indicators into our AI models to foresee sector-wide layoffs, further protecting our portfolio from macro-concentration risks while pre-approving relief structures for affected workers.