Fintech · BWT Hackathon · Team CodeCorps
AI that explains the risk, and never invents the numbers.
An AI-powered Financial Stability Companion for low-income households (around ₹15,000 a month) making safer money decisions. The backend computes every number deterministically, and the AI acts strictly as an interpretation layer on top, so its advice stays grounded in the user's own data.
WebGL · move your cursor
0
Deterministic metrics
0
Risk levels
0
API endpoints
0
MongoDB collections
// highlights
Whatmakesittick
AI as interpreter, not calculator
All financial figures come from a deterministic metrics engine. The LLM only explains them, strictly grounded in the user's data, which keeps advice from drifting into guesses.
Financial metrics engine
Computes Debt-to-Income Ratio, Emergency Fund Coverage and Savings Rate from the user's profile and logged expenses.
Decision evaluation
Rule-based risk assessment for big moves such as taking on a new EMI, returning Low, Moderate or High based on current financial health.
Expense tracking
Daily expenses with categorised summaries and disposable-income monitoring.
Data-grounded chat
A conversational assistant that explains the risk behind a decision in the context of the user's own numbers, with history kept per user.
Secure by default
Clerk handles sign-in and sessions on the front end, and a middleware on the FastAPI side verifies the Clerk JWT on every endpoint.
// architecture
Howit'sbuilt
- 01React + ViteDashboard · expenses · decisions · chat · Clerk
- 02FastAPIMetrics engine · rule-based risk · AI service
- 03MongoDB AtlasUsers · expenses · decisions · chat messages
step 01
Sign in
Clerk authenticates the user and the backend verifies the JWT.
step 02
Record
Income, fixed expenses, EMIs and daily spending are stored per user in MongoDB.
step 03
Compute
The metrics engine derives debt-to-income, emergency-fund coverage and savings rate.
step 04
Evaluate
A rule-based check rates a proposed decision Low, Moderate or High risk.
step 05
Explain
The AI turns those numbers into a plain-language explanation.
// under the hood
Thedetails
API overview (Clerk JWT required)
| Module | Method | Endpoint |
|---|---|---|
| Expenses | POST / GET | /expenses |
| Expenses | DELETE | /expenses/{id} |
| Decisions | POST | /evaluate-decision |
| Decisions | GET | /decision-history |
| AI chat | POST / GET | /chat · /chat-history |
Built for low-income households, and what's next
- ◆Entered in the Future Finance Platforms track under the Financial Clarity problem statement.
- ◆Planned: what-if simulations, credit risk scoring, regional language support, microfinance integration and a PWA for low-end phones.
// stack
Builtwith
- React
- Vite
- Tailwind CSS
- Clerk
- FastAPI
- Python
- MongoDB Atlas
- LLM API