FundIt

Inspiration

Charitable organizations are the cornerstone of a supportive society, playing a critical role in addressing overlooked community needs through collective action and engagement. Many of these organizations rely partially or entirely on a consistent base of donors to sustain their work.

Currently, hours of manual research are spent trying to identify potential donors, and these efforts often fail to lead to contributions due to misaligned interests between donor and applicant.

What if there were a tool that could accurately match organizations with donors who are genuinely likely to support their cause?

What it does

On the front end, the website features an agent-powered chat interface that guides user input, a live dashboard displaying key insights at a glance, and glass-style panels to organize and present information clearly.

On the back end, the system includes a web scraper that collects donation data from relevant websites, a machine learning model that predicts donor engagement based on organization-level and cause-specific metrics, and an LLM-powered intake agent that prompts users to provide key information for predictive donor analysis.

How we built it

First, we trained a predictive model to estimate the likelihood of a donor contributing to a given cause by wrangling donation data to create donor-specific and donor-cause-specific metrics, along with binary labels for each cause. The features we used to predict a likely donor are; the total amount donated, the average amount donated, number of donations, days since last donation, donor type, province, the total amount donated to the cause, average amount donated to the cause, and number of donations to the cause.

Then, we developed a web scraper using BeautifulSoup to extract external donation data from public websites, ensuring the scraped data conformed to the same schema as the internal dataset.

Then, we built an LLM-powered intake agent to conversationally collect user input, specifically the organization’s cause.

Finally, we integrated all components into a pipeline where the web scraper feeds new data into the model, and the LLM provides structured user input for real-time donor-cause compatibility analysis.

What's next for FundIt

Our next steps in development would be to use post-query user data (whether a donor contributed and the donation amount) to refine the predictive model, to add a scrollable donor database with direct links to donor websites integrated with chat results, and to incorporate any user feedback to make the interface more intuitive and actionable.