WiraRakyat

Inspiration WiraRakyat was born from a desire to bridge the digital and physical divide in Malaysian neighborhoods. We were inspired by the resilience of local "Warung" owners and the community spirit during flash floods, realizing that a "uniquely Malaysian" challenge requires a "uniquely Malaysian" AI solution that honors our culture and protects our future.

What it does WiraRakyat is a hyper-local resilience hub that uses Gemini 3.1 multimodal AI to serve three core pillars:

AI Warung Digitizer: Vendors take a photo of their stall or handwritten menu, and Gemini instantly extracts items/prices and generates localized "Manglish" marketing copy to boost their digital presence.

Banjir-Alert Lens: Residents report disaster hazards by snapping photos of rising waters or blocked drains; the AI assesses severity in real-time and triggers a visual "Red Pulse" on a shared 3D map.

Dialect-First Interface: To ensure inclusivity, the entire UI can toggle between English, Malay, and regional dialects like Kelantanese, ensuring that critical safety and business tools reach everyone, including the elderly.

How we built it We engineered a "Solid" and high-performance tech stack designed for "One-Army" solo development:

Frontend: Next.js and React Three Fiber (Three.js) to create an interactive 3D map of Malaysia that acts as the primary navigation hub.

Backend: Supabase handles real-time data synchronization for flood alerts and secure vendor profiles.

AI Integration: We utilized the Gemini 3.1 Flash API for native multimodal processing, allowing the app to "see" and "understand" Malaysian contexts without complex pre-processing.

Deployment: The application is containerized and deployed on Google Cloud Run for maximum scalability and reliability.

Challenges we ran into The primary challenge was optimizing the Three.js 3D map for mobile performance, especially when rendering dozens of "Warung" markers simultaneously. We solved this by implementing InstancedMesh to minimize draw calls. Additionally, training the AI prompt to recognize non-standardized handwritten menus in local contexts required extensive refinement of our Prompt Design.

Accomplishments that we're proud of We are incredibly proud of building a "Triple-Threat" application that successfully integrates economic growth, disaster safety, and cultural preservation into a single, cohesive user experience. Seeing the AI accurately translate a complex disaster hazard into a simple, urgent dialect-based warning was a major milestone for our "Building AI responsibly" mission.

What we learned This project taught us the power of multimodal AI in solving "last-mile" problems for non-developers and traditional businesses. We learned that technical integration is most effective when it is invisible—where a user just takes a photo and the AI handles the complexity of data extraction and neighborhood broadcast.

What's next for WiraRakyat We plan to deepen our LearnLM integration to include educational modules that help youth re-learn their local dialects through interactive 3D "Language Orbs". We also aim to collaborate with local municipal councils to integrate our Banjir-Alert data into official state disaster response dashboards for an even smarter, more inclusive future.