Inspiration
The insight that drove everything was this: your tax refund lands on the 15th, your credit card is due on the 18th, and you have a flight booked for the 20th — a good debt plan has to know all three of those things before it tells you anything. That's what calendar intelligence means to us, and that's what ClearDebt is built around. We've all been there, staring at multiple debts, unsure which one to pay first, while trying to juggle rent, groceries, and an upcoming expense you almost forgot about. Most financial tools tell you your balance but not what to do about it. We built ClearDebt because we wanted something smarter — a system that looks at your actual debts, your real spending patterns, and your calendar all at once, then uses AI to tell you exactly what to do and when.
What it does
Most debt repayment tools ignore the reality of modern life — that a "Study Group at Coffee Shop" event on your calendar means $25 in spending, or that three social events in one weekend requires proactive budget adjustments. ClearDebt helps users eliminate debt faster by combining real banking data, calendar intelligence, and AI-powered repayment strategies. Instead of generic advice, every recommendation is grounded in the user's actual balances, spending behavior, and upcoming financial events.
Features:
1. Multi-Source Debt Aggregation Connects to real bank accounts via Plaid (credit cards, lines of credit, student loans, auto loans) Manual debt entry for accounts outside Plaid All data unified in Snowflake as a single financial profile
2. Spending Behavior Analyzer Reads 90 days of transaction history and surfaces patterns — average monthly spend by category, recurring subscriptions, and seasonal spikes. This output feeds directly into the AI's reasoning when generating your repayment plan. Shows personalized spending insights - like a warning, win, or tip based on your actual financial patterns. It looks at your real Snowflake data: debts, transactions, and spending habits, and surfaces your single most relevant recent achievement, like a streak, savings goal hit, or budget win, to keep you motivated.
3. Calendar Intelligence Layer Overlays debt due dates, income events, and planned expenses on a monthly calendar Detects cash flow conflicts when debt payments and large expenses fall in the same week Calculates the optimal window to make extra debt payments (after income arrives, before next due date) Flags months where total outgoing payments exceed expected income
*4. AI Event Spending Predictor * The core innovation. When you add a calendar event like "Birthday Dinner Downtown" or "Weekend Trip to Niagara Falls", Gemini AI instantly predicts the likely spend, category, and itemized breakdown — before you even think about it.
5. AI Repayment Plan (Snowflake Cortex) Compares Debt Avalanche (highest APR first) vs Debt Snowball (smallest balance first) Generates a personalized hybrid strategy using Snowflake Cortex (Mistral Large 2) Factors in calendar events and spending behavior — not just interest rates Shows total interest saved, payoff date, and month-by-month balance breakdown
6. AI Financial Chat (Gemini) Context-aware chat grounded in the user's actual debt data and repayment plan Handles hypotheticals: "What if I got a $3k bonus?" or "Should I consolidate?" Full conversation history re-injected on every message for continuity
7. Live Dashboard A real-time overview of debt progress, weekly spending analysis, daily cost distribution, spending categories, upcoming calendar events with predicted costs, and milestone alerts.
How we built it
We built ClearDebt on a Snowflake-centered architecture, using it as both our data warehouse and AI inference layer via Snowflake Cortex. Plaid pulls in real bank accounts and transaction history, which are stored and queried in Snowflake.
Our Python FastAPI backend, hosted on DigitalOcean, handles the core business logic. A deterministic math engine pre-computes avalanche and snowball repayment strategies, then sends those results alongside the user’s debt profile and calendar events to Cortex (Mistral Large 2) for personalized recommendations. The calendar intelligence layer cross-references debt due dates, income events, and planned expenses to detect cash-flow conflicts and determine optimal payment windows.
Google Gemini 2.5 Flash powers the context-aware financial chat assistant, with the full debt and repayment context re-injected on every message to maintain grounded responses.
The frontend, built with TypeScript and React, is hosted on Vercel and uses Google OAuth for authentication, enabling secure user access and seamless integration with the backend APIs.
Challenges we ran into
Sleep deprivation aside, our biggest technical challenge was making the calendar integration genuinely intelligent. — it's easy to display a calendar, it's hard to make it reason about whether your paycheck lands before or after three bills are due and what that means for your debt plan. We also underestimated how much prompt engineering goes into getting AI to give specific, number-grounded financial advice instead of generic tips. And at some point around 3am someone accidentally wiped the database and we had to rebuild it from scratch, which was a character-building experience.
Accomplishments that we're proud of
We're proud of how cleanly the architecture came together: Snowflake sitting at the center handling both data and AI inference kept everything tight. And given that most of this was built on zero sleep in under 24 hours, the fact that it works, looks good, and tells you something genuinely useful about your finances feels like a real accomplishment.
Built With
- fastapi
- gemini
- node.js
- python
- react
- snowflake
- tailwind
- typescript
Log in or sign up for Devpost to join the conversation.