WayFinder

Inspiration

Three years ago, while volunteering for a massive international business conference, I witnessed a recurring problem: attendees were constantly lost. Despite having digital "paper maps," people struggled to navigate complex indoor spaces, missing sessions and losing valuable time. My research showed that existing solutions—especially high-end AR guides—are prohibitively expensive and require extensive hardware like sensors and beacons. This makes them impractical for seasonal events or large-scale public spaces. I sidelined the idea until this hackathon, where I discovered that Gemini’s multimodal features could solve the user’s side of the story without any of the traditional hardware costs.

What it does

WayFinder allows users to stop worrying about the map and focus on their destination. It provides a "zero-friction" onboarding experience where users can set their starting point by simply snapping a photo of a nearby store, using their voice, or searching the directory. Key features include: Accessible Routing: Automatically calculates the best path for all user needs (e.g., avoiding stairs). Multimodal Guidance: Users can toggle between an AR overlay for visual direction, a synced audio guide, or a traditional 2D progress map. Interactive Path Confirmation: At key milestones, the app provides visual cues and confirmation buttons to ensure the user stays on track. Flexible Navigation: Users can seamlessly switch between guidance modes mid-journey if their environment or preferences change.

How we built it

I developed WayFinder through a rigorous, iterative process that combined ethnographic research with cutting-edge multimodal AI. User Persona & Strategy: I began by researching the specific needs of mall shoppers and conference attendees, using Gemini 3 Flash to help define user personas and map out "happy paths" for quick navigation versus complex "edge cases" like multi-level floor changes. AI Architecture in Google AI Studio: I utilized Google AI Studio and Gemini 3 Flash to architect the app's reasoning engine, crafting system prompts that allow the AI to act as a spatial interpreter for multimodal inputs—like storefront photos or voice queries. Visual Positioning Logic: I built the system to identify a user's "Start Point" through visual recognition of store signage, leveraging Gemini’s vision capabilities to replace the need for expensive, permanent hardware sensors. Data Sourcing & Grounding: To ensure a "Source of Truth," I initially performed manual scraping of store lists and then progressed to using Firecrawl to aggregate high-fidelity data, including unit IDs like Eataly (C306) and Canada Goose (3110a).

Challenges we ran into

Indoor navigation requires a "Source of Truth" to be reliable. Initially, I experimented with various web-scraping methods to gather store data, but I quickly realized that a direct API connection (like Mappedin) is the only way to ensure accuracy in a dynamic mall environment. Another challenge was the UX of existing mall directories; they often feel like static kiosks ported to a phone. I had to rethink the interface to ensure it wasn't just "tracking" the user, but actively guiding them without requiring them to stare at their screen for the entire walk.

Accomplishments that we're proud of

I successfully built an accessible indoor navigation tool that brings the convenience of outdoor GPS to the most crowded and dynamic indoor spaces—malls, art galleries, and conference centers—without requiring expensive hardware sensors.

What we learned

This project taught me that while the backend of indoor navigation is incredibly complex, the frontend must be invisible. The most powerful solution is one that doesn't ask the user to log in or learn a new system—it just works the moment they need it.

What's next for WayFinder

The goal is to partner with established indoor mapping companies to enhance their consumer-facing interfaces. By combining their robust data with my AI-driven guidance layer, we can finally retire the "digital paper map" for good.

Built With

• capcut
• firecrawl
• gemeni
• google-ai-studio