fundingflow

Funding Library page with real funding programs scraped and parsed with AI
Mock farm profile page
Users funding program page
Real fund program example
Getting recommended funding programs based on user profile and specified problem
AI assisted application planning (1): General information about program
AI assisted application planning (2): Assistance in filling out forms

Inspiration

Farmers struggle to find government funding they qualify for, and existing tools have stale data with no application guidance.

What it does

FundingFlow uses AI to automatically discover funding programs from government websites, match farmers to them via natural language, and generate personalized application plans with auto-filled PDF forms.

How we built it

Node.js/Express backend with Featherless.ai (Qwen3-32B) for AI, Cheerio for web scraping, and pdf-lib for form filling.

Challenges we ran into

Managing token costs from massive government web pages and deduplicating 82+ scraped programs required aggressive HTML cleaning and multi-strategy matching algorithms.

Accomplishments that we're proud of

We built a complete soup-to-nuts workflow from “my tractor broke” to a filled-out government PDF, all powered by four distinct AI applications using real USDA data, not mocks.

What we learned

Prompt engineering varies dramatically by task; extraction needs strict schemas and low temperature, while plan generation needs rich farm context and tier-specific templates.