Tech Stack
Overview
FinTrack is an end-to-end personal finance management system designed to help users track expenses, set budgets, and analyze spending patterns through clean, interactive dashboards. What makes this project unique is how it was built — it’s an experimental exploration of prompt engineering as a development methodology. Instead of traditional coding flow, this project was developed almost entirely through AI-assisted iteration on Replit, using natural language prompts to scaffold unfamiliar frameworks and architecture in TypeScript + React + Node.js. This approach tested how far AI can accelerate full-stack application development for experienced developers exploring unfamiliar stacks.
Objective
To explore how prompt engineering and AI-assisted coding can be leveraged to:
- Accelerate full-stack development in a new tech ecosystem
- Test the boundaries of human–AI collaboration in real software design
- Generate clean, maintainable, production-style code through iterative refinement
Key Features
- Interactive cash flow visualization (income vs. expense trends)
- Real-time budget progress and category breakdown
- Recent transactions list with filtering and search
- Expense trend analysis with customizable time ranges
- CRUD for transactions with categories and tags
- Budget management with progress indicators and thresholds
- Visual cues for overspending and under-budget status
- Reports for income/expense comparison and category insights
- Modularized data visualization layer built with Recharts
- Custom date range filters and performance metrics
- Export-ready structured financial summaries
Technical Highlights
Frontend
Tools & Techniques: React, TypeScript, TanStack Query, ShadcnUI, TailwindCSS
Notes: Modular, component-driven UI with reactive data management
Backend
Tools & Techniques: Node.js, Express, Drizzle ORM, PostgreSQL
Notes: REST APIs with typed schema, structured routing, and clean separation
Infrastructure
Tools & Techniques: Docker + Compose
Notes: Simplified full-stack deployment for reproducibility
AI Methodology
Tools & Techniques: Replit + Chat-based prompt engineering
Notes: Used structured prompts to auto-generate modules, fix errors, and refine UX logic iteratively
Experimental Methodology — Prompt Engineering Approach
Rather than manually building every component, each part of FinTrack was developed through carefully designed prompts, including:
“Generate a responsive budget dashboard with category pie chart using Recharts and Tailwind”
“Create TypeScript Express route for adding transactions using Drizzle ORM”
“Fix Docker Compose setup for client-server with shared schema folder”
This iterative cycle tested the ability of AI tools to:
- Understand multi-layered requirements
- Generate production-quality TypeScript code
- Maintain architectural consistency across frontend, backend, and shared models
Through this process, FinTrack became both a tool and a case study in using AI for rapid prototyping.
System Architecture
client/ → React + TypeScript frontend (modular components)
server/ → Express + Node backend
shared/ → Common schema definitions (Zod + TypeScript)
db/ → PostgreSQL + Drizzle ORM migrations
docker/ → Compose setup for full-stack orchestration
Data Flow:
Frontend (React Query) → API Gateway (Express Routes) → Drizzle ORM → PostgreSQL → Aggregated analytics visualization
Impact & Insights
- Developed a full-stack TypeScript application in under 7 days using AI prompting.
- Achieved feature parity with typical personal finance apps — dashboards, CRUD, and visual analytics — without prior React/TypeScript expertise.
- Demonstrated that AI-assisted coding can reduce onboarding time to new frameworks by over 60%.
- Showcased AI’s current limits — context retention, state dependency, and refactor complexity — providing valuable learnings for future iterations.
Learnings
- Prompt quality determines output quality — specificity matters more than length.
- AI accelerates scaffolding but still needs human architectural judgment.
- AI-driven collaboration is best for rapid prototyping, not yet for large-scale maintenance.
Future Enhancements
- Multi-user authentication (JWT or OAuth)
- Financial goal tracking module
- OCR-based receipt scanning and auto-categorization
- Integration with bank APIs for real transaction imports
- Machine learning–based spending pattern prediction
Project Status
Stage: Experimental / MVP Complete
Platform: Replit (AI-collaborative environment)
Deployment: Docker-ready | Localhost (dev mode) | PostgreSQL integration
“FinTrack wasn’t just a coding project — it was an experiment in how thinking in prompts can replace thinking in syntax. It taught me that curiosity and structure together can make unfamiliar technology approachable, even enjoyable.”
View on GitHub