Ujjwal Kashyap, Software Developer

pAIse is a next-generation financial utility application designed to solve the perennial chaos of shared expenses among friends, roommates, and travel groups. While traditional apps like Splitwise act as simple ledgers, pAIse leverages Generative AI to act as an intelligent financial assistant, automating data entry and providing contextual insights into user spending.

The Problem & Solution

Managing shared finances is often tedious, requiring manual data entry, complex math for unequal splits, and awkward reminders. pAIse ("pay-ease") eliminates this friction by treating expense tracking as a conversation rather than a data-entry task. Users can interact with the app naturally—through voice commands, images, or text—while the system handles the complex logic of debt simplification and balance tracking in the background.

Key Innovations & AI Integration

The core differentiator of pAIse is its deep integration of Google’s Gemini AI models to streamline user interaction:

• Voice-Powered Entry: Users can simply speak naturally (e.g., "I paid 500 for lunch with Bob, split equally"). The app captures the audio, transcribes it, and extracts structured data to create the transaction automatically.
• Smart Receipt Scanning: An integrated OCR system analyzes photos of physical receipts, identifying line items and totals to populate expense forms instantly.
• Context-Aware AI Chatbot: A dual-mode chatbot serves as both a support agent and a financial analyst. It uses RAG (Retrieval-Augmented Generation) to answer feature queries and securely accesses personal financial contexts for authenticated users.

Technical Architecture & Cloud Infrastructure

pAIse is architected for performance, scalability, and security using a modern serverless stack deployed on Free Tier services to showcase high-value engineering with zero initial cost:

• Frontend & Edge: Built with Next.js 14 (App Router) and TypeScript, deployed on Vercel's Free Tier, leveraging the global Edge Network for low-latency performance.
• Database & Storage: The backend relies on Supabase (PostgreSQL). It provides robust relational data persistence, utilizing CUIDs for security and complex schemas for debt handling.
• Caching & Rate Limiting: Upstash Redis is implemented at the edge. It effectively handles caching and strictly enforces rate limits, protecting the AI endpoints from abuse and DDoS attacks.
• Security-First Design: Implements a "Defense-in-Depth" strategy, including strict HTTP Security Headers (HSTS, CSP), Zod validation schemas, and secure, HTTP-only cookies.

Impact

pAIse transforms a mundane administrative chore into a seamless, intelligent experience. By combining rigorous full-stack development practices with a sophisticated, zero-cost cloud infrastructure (Vercel, Supabase, Upstash), this project showcases my ability to build secure, scalable, and user-centric software that solves real-world problems effectively.