Nekofi

full features

Inspiration

Traditional budget trackers usually fall into two categories: they either force you to manually input and categorize every single transaction like a glorified spreadsheet, or they require you to connect your bank accounts to third-party servers, compromising your privacy and data security. Our inspiration was to build a smart, playful, and completely private personal finance companion. We wanted the intelligence of modern AI to do the heavy lifting of categorizing expenses and providing insights, but we wanted to do it strictly on-device so your sensitive financial data never has to leave your phone.

What it does

Nekofi is a privacy-first, AI-powered personal budget tracker that lives on your mobile device. It allows users to effortlessly track their daily expenses, visualize their spending habits with beautiful interactive charts, and receive intelligent financial insights. Beyond individual tracking, Nekofi features collaborative shared budgets, allowing you to easily invite family members, partners, or roommates to manage joint expenses and track shared financial goals together. What sets Nekofi apart is its on-device AI assistant—it automatically categorizes your transactions, detects spending anomalies, and answers conversational questions about your budget using a local language model. You get all the benefits of a "smart" financial advisor and seamless collaborative tracking without sacrificing your sensitive data to the cloud.

How we built it

Frontend: **We built the mobile application using React Native and Expo (SDK 51), ensuring a fluid, native-like experience across platforms. We used Zustand for lightweight and fast state management. **Backend: We utilized Supabase for our backend infrastructure, taking advantage of its PostgreSQL database for robust data syncing, Supabase Auth/JWT for secure user authentication, and Supabase Storage. Local AI: The core of our privacy-first intelligence is powered by llama.rn, which allowed us to run a Large Language Model (LLM) completely locally on the user's device. Data Visualization: We integrated React Native Chart Kit to render clean, responsive graphs that help users visualize their financial health at a glance.

Challenges we ran into

Running a Large Language Model directly on a mobile device was by far our biggest hurdle. We had to carefully balance the model's parameter size with the device's memory constraints to ensure it didn't crash the app or excessively drain the battery. Integrating llama.rn into the React Native ecosystem required navigating complex native build configurations and managing asynchronous tasks so the UI wouldn't freeze while the AI was "thinking." Additionally, designing an intuitive user interface that could display complex financial data while seamlessly incorporating conversational AI responses took several design iterations.

Accomplishments that we're proud of

We are incredibly proud of successfully deploying a local LLM on mobile to analyze personal finances without compromising user privacy. Achieving completely offline, real-time AI expense categorization that feels snappy and accurate is a massive technical milestone for us. Furthermore, merging this bleeding-edge AI capability with a polished, premium UI and a robust cloud-syncing backend (via Supabase) makes Nekofi feel less like a hackathon prototype and more like a production-ready product.

What we learned

This project pushed us to the bleeding edge of mobile AI. We learned a tremendous amount about optimizing local LLMs for resource-constrained environments, dealing with memory overhead, and working with complex native modules in React Native. We also deepened our understanding of the Supabase ecosystem, particularly in structuring secure PostgreSQL schemas and utilizing JWTs to maintain a seamless, authenticated connection between the local app and the cloud.

What's next for Nekofi

Moving forward, we plan to implement local receipt scanning using the device camera and offline OCR, allowing users to log expenses simply by snapping a photo. We are also exploring the possibility of fine-tuning a smaller, highly specialized financial model to make the on-device AI even faster and more accurate at parsing transaction details. Finally, we want to leverage Supabase's real-time capabilities to introduce secure collaborative budgeting, so families or partners can manage joint finances together.