Aegis

Our journey didn't start with a clear roadmap; it started with a blank canvas and a deep dive into the hackathon tracks. We noticed that Databricks highlighted a significant challenge regarding large-scale data processing for social impact. This sparked an idea: could we create an AI-driven "shield" for the world’s most vulnerable? We realized that while global attention shifts rapidly, many humanitarian crises suffer in silence. We founded Aegis to bridge this gap, using predictive intelligence to identify exactly where the world is looking away and quantifying the "underfunding gap" that leaves millions without support.

Aegis was engineered as a dual-engine system designed to turn raw humanitarian metrics into actionable relief strategies. The core of our project is a predictive model built to track funding trends and calculate the disparity between a population's actual needs and the aid they receive. We utilized Python and Databricks to process historical data on population, GDP, and crisis type. By applying linear programming and similarity searches, we developed a system that doesn't just identify a deficit—it suggests a roadmap. The "Solution Engine" matches current crises to historical success stories, simulating how an optimal allocation of "full funding" could be distributed across sectors like health, shelter, and food to maximize life-saving impact.

Our biggest hurdle was the "Data Desert." When we first began, we hit a wall trying to gather clean, granular humanitarian datasets for every specific region. We quickly realized that the data we wanted didn't always exist in a ready-to-use format. To overcome this, we had to innovate: we used data to predict other data. We pivoted our strategy to analyze the relationship between "Requested Funds" and "Committed Money." By training our model on these historical ratios, we were able to synthesize the missing pieces of the puzzle. This allowed us to project the underfunded amount even when primary data was sparse, turning a potential project-stopper into our most unique technical feature.

This project was a masterclass in Bounded Rationality and engineering resilience. We learned that in a hackathon—just like in real-world humanitarian work—you rarely have perfect information. We gained deep experience in data synthesis and learned how to leverage the Databricks ecosystem to handle complex, multi-layered datasets. Most importantly, we learned that the most powerful AI models aren't just the ones with the most data, but the ones that can intelligently navigate the gaps in that data to find the truth.

Built With

• css
• flask
• groqai
• node.js
• tailwindcs
• typescript
• vite