Runway

Inspiration

I’m a startup founder, and like many first-time founders, I’ve realized how little financial literacy I actually have when it comes to running a business. Tracking runway, understanding unit economics, and modeling decisions like hiring or marketing spend felt overwhelming without a CFO.

That made me wonder: If I, with access to resources, struggle with this — what about micro-entrepreneurs like food truck owners, artisans, or gig workers who don’t have tools, training, or time to analyze their finances?

Runway was inspired by that gap: to create a financial co-pilot that makes business decision-making clear, simple, and accessible for every entrepreneur.

What it does

Runway is an AI-powered financial co-pilot designed specifically for micro-entrepreneurs. It provides a mobile-first financial account that goes far beyond basic banking for consumers.

Intelligent Ledger: Automatically connects to existing payment accounts (Plaid sandbox for demo) and uses AI to categorize every transaction into clear business buckets like Revenue, Inventory, Marketing, and Operations. This instantly creates a real-time, easy-to-understand Profit & Loss statement.

Causal Decision Engine: It allows entrepreneurs to simulate the financial impact of key business decisions, like "What if I hire a part-time assistant?" or "What if I invest in a new espresso machine?", before they commit any capital. It provides a visual cash flow forecast, comparing their current trajectory against their proposed future.

Automated Stakeholder Reports: Generates plain-language summaries of monthly business performance. These reports empower users with low financial literacy to track their progress and communicate effectively with family, community lenders, or potential investors.

How we built it

Frontend: React Native for a mobile-first experience. Backend: FastAPI (Python). Data Layer: Mock transaction datasets modeled after Plaid/Stripe output. Machine Learning: Hugging Face zero-shot classifier for transaction categorization; Prophet for forecasting cash flows; Generative AI API (Gemini) for plain-language reports. Database: Supabase + PostgreSQL for storage and user management.

Challenges we ran into

Plaid Integration & Sandbox Data: Integrating with Plaid in a hackathon setting required careful handling of authentication flows and understanding the nuances of sandbox data to simulate realistic transaction streams.

ML Model Configuration for Categorization: Fine-tuning the zero-shot classifier to accurately categorize diverse, often ambiguous transaction descriptions (e.g., 'SQ *STARBUCKS' vs. 'SQ *HOME DEPOT') into business-specific buckets required iterative testing and clever prompt engineering.

Real-time Forecasting with Prophet: Ensuring the Prophet model could generate and update cash flow forecasts quickly for instantaneous scenario simulation, especially with limited initial data, was a performance optimization challenge.

Simplifying Complex Financial Concepts: The biggest UX challenge was translating sophisticated financial metrics and the output of our AI models into an interface and language comprehensible to users with low financial literacy.

Accomplishments that we're proud of

Fully Functional AI Core: I successfully implemented and integrated three distinct AI models, a transaction categorizer, a powerful cash flow forecaster, and a generative report writer; all working seamlessly together to provide intelligent financial assistance.

This is it really. I've never had a solo hackathon experience, so building a product I'm happy to present to people is a huge milestone for me.

What we learned

One thing is that having more hands to work is not necessarily better. Although the diverse perspectives help, there may be conflict. By working solo, it's on my own responsibility to see the event through.

Two other things I want to say is (1) by deeply understanding the specific financial challenges of micro-entrepreneurs, I felt that I was able to design a solution far more impactful than a generic financial app. (2) Designing an interface for users with varying degrees of financial and digital literacy required constant empathy and iteration to ensure clarity and simplicity. Things got too complicated at times and I had to dial it down.

What's next for Runway

A lot. Integration with real APIs (Plaid, Stripe, Cash App) so entrepreneurs can onboard instantly. Multi-language support to make reports accessible to more communities. Credit-building and micro-loan partnerships, so our generated reports can help entrepreneurs access growth capital. Community benchmarking: comparing costs and margins against anonymized peers (e.g., “your food costs are 10% higher than average taco trucks in Houston”). AI business coach: a conversational interface where entrepreneurs can ask “Can I afford to hire?” and get scenario-backed answers.

Built With

• fastapi
• gemini
• hugging-face
• plaid
• postgresql
• prophet
• python
• react