Cashly

Inspiration

Have you ever tried to manage your money as a student but felt completely overwhelmed? Tracking spending, setting goals, and planning investments feels complicated, and most finance apps are either dry, robotic, or demand manual entry. Cashly solves this by combining secure, real-time bank integration through Plaid with LLaMA AI, delivering personalized, humorous, and student-friendly financial advice. It provides a holistic view of your finances, spending, goals, and investments, all in one place; Plaid makes money management effortless, insightful, and even a little fun.

What it does

Cashly is a personal finance platform designed specifically for students, making money management simple, insightful, and even fun. It tracks spending, income, and investments automatically, giving students a clear picture of where their money is going.

Cashly categorizes transactions into intuitive groups like food, entertainment, and transportation, highlights trends over time, and helps users spot habits or areas where they can save. Students can set financial goals, such as saving for a laptop, building an emergency fund, or planning a trip, and visually track their progress with helpful insights along the way.

Investments can also be monitored, showing portfolio performance and helping students learn the basics of investing without feeling overwhelmed. AI financial insights are given.

Most uniquely, Cashly uses AI to generate personalized, humorous advice tailored to each user’s spending behavior. It offers practical recommendations while keeping the tone light and engaging. For example, suggesting ways to cut back on small recurring expenses or congratulating users on hitting a savings milestone.

Cashly transforms financial management from stressful and confusing into effortless, intuitive, and even enjoyable, helping students make smarter financial decisions and achieve their goals.

How we built it

Frontend: React with TypeScript, Tailwind CSS for styling, Recharts for charts, Framer Motion for smooth animations, and Radix UI for accessible components.

Backend: Convex handles real-time updates and authentication. Python services process PDFs (via PDFPlumber) and generate AI insights using LLaMA through Ollama.

Integrations: Plaid securely fetches live bank data, and Alpha Vantage provides real-time stock prices.

Data Flow: Bank accounts/PDFs/manual input → transactions categorized → AI analyzes patterns → dashboards, goals, and investments update instantly.

Key Features: Real-time updates, scalable microservices, responsive design, secure data handling, and student-focused usability.

Challenges we ran into

PDF Processing: Extracting transactions from diverse bank statements was tricky. PDFs vary in format, so we used PDFPlumber with Python to reliably parse them.

Local Development with AI & PDF Services: To test PDF processing and AI insights in real time while developing locally, we had to use ngrok to expose our local server securely to the frontend. This took a good minute.

Bank Integration: Handling Plaid OAuth flows securely and ensuring real-time updates required reading a lot of documentation.

Data Accuracy & Validation: Ensuring uploaded PDFs matched the user’s account data, and that AI categorizations were correct, required multiple validation layers.

Accomplishments that we're proud of

Seamless Bank Integration: Successfully connected real student bank accounts via Plaid for secure, real-time transaction tracking.

Intelligent AI Insights: LLaMA generates personalized, humorous financial advice that’s both actionable and engaging for students.

Reliable PDF Processing: Built a system with PDFPlumber to extract and categorize transactions from varied statement formats.

Scalable Architecture: Real-time updates, AI processing, and integrations all work smoothly thanks to a modular, microservices-based stack.

What we learned

AI Prompt Engineering: Crafting LLaMA prompts to balance humor, tone, and actionable financial advice was key for student engagement.

PDF Variability: Bank statements come in many formats, so handling this with PDFPlumber taught me data extraction techniques.

Secure Integrations: Managing Plaid OAuth and real-time banking data emphasized the importance of security and privacy in fintech.

Local Dev Challenges: Using ngrok to expose local services for AI and PDF processing taught practical solutions for testing connected systems.

What's next for Cashly

Cashly aims to expand its AI capabilities to provide smarter, predictive insights and more personalized financial advice for students. We plan to add credit score monitoring, bill reminders, group budgeting features, and support for crypto and retirement planning, all while maintaining a fun, friendly experience. Scalability is a key focus. Our modular, microservices based architecture will allow us to support more users, integrate additional banking systems, and handle larger volumes of transactions without compromising performance. We also aim to enhance mobile functionality, improve PDF processing for varied statement formats, and introduce gamification features to keep our users engaged and motivated in managing their finances.