Platform Showcase
A production-grade real estate platform built end-to-end for the Sri Lankan market
Sri Lanka's first AI-powered trilingual property listing platform — built from zero to deployment. Featuring a conversational AI agent, voice I/O, market insights, 3D hero scenes, and full i18n support across English, Sinhala, and Tamil.
A production-grade real estate platform built end-to-end for the Sri Lankan market
A production-grade real estate platform built end-to-end for the Sri Lankan market
Sri Lanka lacked a modern, AI-powered property platform that serves all three language communities. Existing platforms had no intelligent search, no AI assistance, and poor mobile experiences — leaving a gap for millions of users.
Built InnaThanak from the ground up — a Next.js 14 full-stack platform with Supabase backend, Gemini AI conversational agent (Sara), trilingual i18n, 3D immersive UI, and a complete admin dashboard for content moderation.
Deployed to production on Vercel covering all 25 Sri Lankan districts, enabling property listing via AI conversation, voice input, real-time market analytics, and seamless trilingual experiences.
What makes InnaThanak stand out
Conversational AI built with LangChain + Google Gemini that guides users through property listing creation step-by-step via text, voice, and image input.
Web Speech API integration with intelligent voice selection per language — speak in Sinhala, Tamil, or English and Sara responds accordingly.
Real-time market analytics powered by Gemini 2.0 Flash — investment scores, trend analysis, demand levels, and actionable recommendations.
Full i18n with next-intl across English, සිංහල, and தமிழ் — including property data, AI responses, and all UI elements.
Advanced filtering by type, category, district, price range, bedrooms — with server-side pagination, sorting, and view tracking.
Three.js hero scene with React Three Fiber, Framer Motion page transitions, GSAP parallax effects, and Lenis smooth scrolling.
Guided listing creation with Zod validation, multi-image/video uploads to Supabase Storage, and real-time preview.
Full moderation panel — approve/reject listings, toggle featured status, add admin notes, filter by status, and manage all content.
Supabase Auth with email/password + Google OAuth, Row-Level Security, 4-tier token fallback, and role-based access control.
How the pieces fit together — from frontend to AI
Next.js 14 App Router with SSR/SSG, Tailwind CSS design system, Zustand state management, React Hook Form + Zod validation, and Framer Motion animations throughout.
Supabase PostgreSQL with Row-Level Security, Prisma ORM for GraphQL layer, Apollo Server 5 endpoint, REST API routes for CRUD operations, and Supabase Storage for media handling.
LangChain orchestration with Google Gemini for the Sara conversational agent, Gemini 2.0 Flash for market analysis, structured JSON output parsing, and graceful fallback strategies.
Production-grade tools powering InnaThanak
What I built across the entire platform
3D hero scene, search bar, featured properties, category browsing, popular districts, AI CTA, and floating chat widget.
Full search with filters, sorting, server-side pagination (12/page). Detail pages with image gallery, lightbox, contact options, similar properties.
4-step wizard form (Type → Location → Details → Photos) with Zod validation and multi-file uploads. Full edit capabilities for owners.
Full-page conversational interface with Sara — create listings by chatting, upload photos mid-conversation, with animated voice orb.
AI-analyzed dashboard with stat cards, category breakdowns, trend badges, investment meter, district heatmap, and recommendations.
User dashboard for managing ads, saved properties, profile settings. Admin panel for moderation, approvals, and content management.
Full CRUD for properties, file uploads, AI chat endpoint, market insights API, admin operations, OAuth callback, and Apollo Server GraphQL endpoint.
Email/password + Google OAuth, email confirmation flow, session management with 4-tier token fallback, and auto profile creation via DB trigger.
JSON-LD structured data, dynamic sitemap (all pages × 3 locales), hreflang tags, OpenGraph metadata, and installable PWA manifest.
PostgreSQL schema with Row-Level Security on Supabase
User profiles linked to Supabase auth — name, email, phone, avatar, role (user/admin), preferred language, and timestamps. Auto-created via database trigger on signup.
Core listings table with trilingual titles/descriptions, type (Sale/Rent/Dorm), category, pricing, location (district + city), specs (bedrooms, area), status workflow, view tracking, and AI-posted flag.
Separate tables for property images and videos with sort ordering, plus a saved_properties junction table for user bookmarks. All media stored in Supabase Storage buckets.
Built from zero to production deployment
From concept to deployment — let's bring your platform to life with AI and modern web tech.