Annapurna

Inspiration

👨‍🌾 Inspiration

When I was young, I used to watch my grandfather work tirelessly on his farm. He would pour his heart into growing vegetables, only to watch a significant portion of his harvest rot and go to waste simply because transport trucks broke down or their refrigeration failed on the highway. It broke my heart to see his hard work turn into literal garbage before it ever reached a plate.

Today, up to 40% of food produced in emerging markets is wasted in transit. I built Annapurna to ensure that no farmer ever has to watch their harvest go to waste again.

🚚 What it does

Annapurna is a next-generation, AI-driven logistics platform designed specifically to eradicate food waste in the cold chain. We don't just track trucks; we protect perishables.

Using AI, Annapurna acts as a predictive Fleet Command Center, monitoring environmental telemetry and vehicle health to predict anomalies.

When an unavoidable breakdown occurs (like a compressor failure causing a temperature spike), the AI triggers an Emergency SOS and instantly spins up a geo-fenced Emergency Wholesaler Marketplace. Instead of rotting on the highway, the stranded cargo is pushed to local wholesalers who can place bids on the distressed food at a discount. Wholesalers get turn-by-turn navigation directly to the broken-down truck to rescue the cargo.

The economic recovery model we achieve can be represented by maximizing the salvaged value: $$ V_{salvaged} = \sum_{i=1}^{N} (B_i \times Q_i) - C_{logistics} $$ (Where \( B_i \) is the accepted bid price, and \( Q_i \) is the quantity rescued by local wholesalers).

⚙️ How we built it

• Frontend & UI: Built on Next.js for blistering fast performance and a beautiful, Apple HIG-inspired interface using Tailwind CSS.
• Backend & Realtime: We used Supabase (PostgreSQL) and relied heavily on its WebSockets for live Real-time Streaming. This is how wholesalers instantly see a truck break down on the map without ever refreshing the page.
• AI Brain: Powered by cutting-edge LLMs via Groq for instantaneous anomaly detection and intelligent rerouting.
• Analytics & UX Optimization: Integrated Novus.ai (powered by Pendo) directly into our root layout to track user journeys, optimize the wholesaler bidding flow, and ensure the app adapts to the way drivers actually use it.
• Deployment: Hosted on Vercel for instant global edge delivery.

🚧 Challenges we ran into

The hardest part was building the live bidirectional marketplace. Synchronizing a truck breaking down in the Fleet Manager dashboard and instantly popping up a notification on the Wholesaler dashboards required highly optimized Supabase Realtime subscriptions.

Additionally, configuring Next.js environment variables for a multi-deployment setup and seamlessly injecting the Novus.ai tracking snippet into our React Server Components without breaking hydration was a massive learning curve!

🚀 What's next for Annapurna

We plan to introduce automated payment escrow via Stripe, so wholesalers can securely pay for the distressed cargo before they even arrive at the truck's coordinates. We also want to feed our Novus.ai analytics data back into our AI models to predict where wholesalers are most likely to buy distressed cargo, allowing us to route trucks closer to high-demand zones preemptively!

Built With

• groq
• next.js
• novus.ai
• postgresql
• supabase
• tailwindcss
• typescript
• vercel