Fishlo
An AI-powered B2C fish marketplace modernizing traditional wet markets through real-time negotiation and hyper-local geospatial logistics.
About This Project
What I built and why
Fishlo is an AI-driven B2C marketplace focused specifically on the fish trade, designed to bridge traditional wet markets with modern digital commerce. The platform introduces an AI bargaining assistant, "Meena Tai," which negotiates fish prices with customers in natural language while strictly enforcing business rules and profit margins. To preserve freshness and delivery accuracy, Fishlo leverages GeoDjango and PostGIS for precise, coordinate-based service zoning, dynamic pricing, and location-aware logistics. By combining conversational AI with advanced geospatial intelligence, Fishlo delivers fair pricing, accurate delivery validation, and a secure checkout experience tailored to the fish supply chain.
Key Features
What makes this project stand out
AI Bargaining Assistant (Meena Tai): A natural-language AI agent that negotiates fish prices in real time while enforcing minimum profit margins and business constraints.
PostGIS-Based Geofencing: Polygon-based service zones that restrict fish orders to precise, deliverable areas to ensure freshness.
Dynamic Zone Pricing: Automatic adjustment of fish and delivery prices based on the user’s exact geographic location.
Admin Logistics Dashboard: A map-driven interface that allows administrators to draw, update, and manage fish delivery zones in real time.
Secure Payment Flow: AI-negotiated fish prices are securely locked and synchronized with the Razorpay checkout to prevent price tampering.
Challenges & Solutions
Problems I faced and how I solved them
Challenges Faced
Designing a non-deterministic AI negotiation system that consistently respects dynamic price floors and fish pricing rules.
Synchronizing AI-negotiated fish prices with a rigid financial checkout system without inconsistencies or tampering.
Replacing traditional zip-code delivery validation with high-precision, hyper-local geospatial eligibility checks for fish delivery.
Optimizing real-time spatial queries for high-traffic scenarios involving complex polygonal delivery zones.
Balancing conversational AI workloads with heavy geospatial computations while maintaining application performance.
My Solutions
Developed a custom Negotiation State Machine in Django to securely validate AI responses before committing fish price changes.
Implemented a cryptographically hashed session mechanism that binds the negotiated fish price to a transaction ID before Razorpay checkout.
Integrated GeoDjango and PostGIS with real-world polygon data to ensure accurate fish delivery eligibility determination.
Optimized spatial queries using PostGIS R-tree indexing to enable fast Point-in-Polygon lookups with sub-second response times.
Architected an event-driven backend with Redis caching to decouple intensive geospatial calculations from the AI negotiation flow.