ClearPath

UI
Programs
About

Inspiration

Navigating public benefits, social assistance, and institutional aid programs is notoriously overwhelming. Vulnerable families and individuals frequently face a wall of dense regulatory terminology, disjointed application sites, and conflicting eligibility constraints. This administrative friction causes thousands of people to abandon their search, completely missing out on vital resources. We built ClearPath to substitute complex bureaucratic mazes with an empathetic, streamlined conversational experience that guides people straight to the support they need.

What it does

ClearPath acts as an intelligent, conversational pre-screening navigator for public benefit programs. Instead of making users read through endless pages of legal guidelines, ClearPath takes a user's organic situational details in plain text, processes them using structured workflows, and immediately generates an actionable checklist of benefits they may qualify for, complete with the clear, step-by-step documentation required to apply.

How we built it

ClearPath is structured as a responsive, lightweight web application built on top of a specialized AI agent workflow:

The Backend Engine: Powered by Python and Flask, managing secure user sessions, context windows, and application operational state.
The Core Intelligence: Integrated directly with the Google Gemini API (gemini-2.5-flash) to safely parse conversational intent and check it against complex program parameters.
The Interface: A clean, intuitive frontend styled with HTML5, CSS3, and dynamic JavaScript logic to deliver instant, easy-to-read assistance checklists and guidance.

Challenges we ran into

Managing strict contextual guardrails was our biggest technical challenge. In social support applications, a generic AI can easily give speculative, incorrect, or binding declarations about eligibility. We solved this by implementing an engineering protocol that enforces objective framing. The AI acts exclusively as a pre-screening and informational aid, using tentative tracking language ("You may qualify") rather than making definitive promises, ensuring human agency is preserved.

Accomplishments that we're proud of

We are incredibly proud of building a functional, end-to-end prototype that handles complex multi-turn conversational data smoothly. Successfully decoupling sensitive user intake parameters from permanent databases—relying instead on ephemeral Flask context storage—allowed us to create a privacy-first application without compromising the continuous user experience.

What we learned

Building ClearPath showed us how critical proper state management and system prompts are when constructing AI workflows. We learned how to tightly scope a language model's operational boundaries using behavioral guardrails, ensuring that the assistant remains factual, secure, and focused exclusively on navigation assistance.

What's next for ClearPath

Moving forward, we want to expand ClearPath to include automated document parsing. This would allow users to upload their anonymous utility statements or tax summaries so the platform can automatically pre-fill eligibility tracking checklists, reducing manual data entry burden down to near zero.