Service
Full-Stack AI Platform Development
Tech Stack
Next.js 14, Google Gemini 2.0 Flash API, Prisma ORM, MongoDB Atlas, Cloudinary CDN, TypeScript, NextAuth.js, Recharts, Vercel
Live Project
VisitPredictye AI: AI-Assisted Resale Valuation Platform
Built at Itigai, Predictye is an AI-assisted platform that estimates resale value for electronics, furniture, and jewelry through conversational and image-assisted interactions. This project was a deep dive into LLM-powered reasoning pipelines and building intelligent, conversational user experiences.
Technical Challenge
People struggle to accurately price items for resale either leaving money on the table or overpricing and not selling. The goal was to combine visual AI analysis with conversational UX to provide accurate, contextual estimates.
AI Integration Journey
API Integration: Google Gemini 2.0 Flash for multi-modal AI (text + image processing)
Image Processing: Cloudinary for upload, optimization, and analysis pipeline
Conversation Flow: Built dynamic questionnaire system that adapts based on item category
Data Visualization: Recharts for displaying price trends and confidence ranges
Technical Stack
Frontend: Next.js 14 with App Router for optimal SEO and performance
Backend: Prisma ORM with MongoDB Atlas for flexible data modeling
Authentication: NextAuth.js with Google OAuth integration
AI Services: Google Gemini API for intelligent image and text analysis
Deployment: Vercel with environment-based configuration
Key Features Built
Multi-modal AI analysis combining image recognition with conversation context
Dynamic form generation that adapts questions based on detected item categories
Real-time price estimation with confidence intervals and market trend data
Conversation memory for context-aware follow-up predictions
Save and compare functionality for tracking multiple items
Learning Experience
This project taught me AI API integration, building conversational interfaces, and creating intelligent user flows. The biggest challenge was designing prompts that consistently returned structured data while maintaining natural conversation flow.
Current Status
Live platform processing image uploads and generating AI-powered resale estimates.
Results & Impact
Designed and developed AI-assisted resale valuation workflows for electronics, furniture, and jewelry through conversational and image-assisted interactions
Integrated LLM-powered reasoning and recommendation pipelines to generate contextual valuation insights
Built backend APIs and orchestration services supporting scalable, AI-driven assessment experiences
Developed multi-year value projection system (2-10 years) with confidence intervals and market trends
Created a dynamic questionnaire system adapting to item categories with an extensible architecture