FinTrack-AI : From Excel Sheets to AI Financial Companion

A developer built FinTrack AI, a local-first personal finance tracker that ingests existing Excel and CSV files and lets users query their transactions in natural language. The Streamlit application uses Python, Pandas, SQLite and openpyxl for data processing and storage, and runs Gemma 3 1B locally through Ollama so financial data never leaves the machine. The developer said the goal was to fit into an existing spreadsheet workflow rather than replace it: "Take the Excel sheets they already use and make them much easier to understand.

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend https://dev.to/challenges/hacktoberfest-weekend-2026-10-01 What I Built I built FinTrack AI for my father and uncle. ❤️ My father and uncle manage their financial records using Excel sheets. While Excel is great for maintaining records, manually entering transactions, organizing data, calculating summaries, and trying to understand spending patterns can become repetitive and time-consuming. So I wanted to build something that would work with the way they already manage their finances, instead of making them learn an entirely new system. That's how FinTrack AI was born. FinTrack AI is a local-first AI-powered finance tracker where users can simply upload their existing Excel or CSV financial data and start working with it. Instead of manually going through rows and formulas, they can: 📂 Upload their existing Excel/CSV files 📊 Automatically organize and analyze their transactions 💰 View financial summaries and spending insights 📈 Explore their expenses through a dashboard 🤖 Ask questions about their financial data using AI 🧠 Get AI-generated explanations and summaries 📥 Export their processed financial data back to Excel 💾 Keep their financial information locally For example, instead of manually searching through an Excel sheet, they can ask questions such as: "Where did I spend the most this month?" "What are my biggest expense categories?" "Give me a summary of my spending." "What are my major transactions?" The goal is simple: Take the Excel sheets they already use and make them much easier to understand. I didn't want to build another complicated finance application that required them to completely change their workflow. I wanted to build something that fits into it. Upload → Analyze → Ask → Understand → Export 🎥 Demo Video: Video Link https://drive.google.com/file/d/1Hsjiws5SpcCAVl7NsJv-I-ZJJmU5FfEk/view?usp=sharing The demo will show the complete workflow of FinTrack AI: Upload an existing Excel/CSV file Import and process the financial transactions View the financial dashboard and summaries Ask FinTrack AI questions about the data Receive an AI-generated explanation Export the financial data back to Excel 💻 GitHub: Github Repository https://github.com/vaishnavipophalkar97-cmyk/FinTrack-AI The complete source code is available in the repository, including the Streamlit application, financial data processing, SQLite database integration, Excel/CSV handling, and local Gemma + Ollama integration. FinTrack AI is built as a local-first financial application, combining traditional data processing with an open-weight AI model. 🛠️ Tech Stack Python — core application logic Streamlit — user interface SQLite — local financial data storage Pandas — financial data processing and analysis openpyxl — Excel file handling Ollama — local AI model runtime Gemma 3 1B — open-weight AI model 🤖 How Gemma is actually used The AI component is powered by Gemma 3 1B, running locally through Ollama. FinTrack AI does not simply send the user's financial information to a cloud AI API. Instead, the application runs Gemma locally through Ollama. The local inference flow is: User's Financial Data → Python/Pandas Analysis → Relevant Financial Context → Gemma 3 1B → Natural-Language Explanation Python and Pandas handle the actual numerical processing and financial calculations. Gemma's role is to take the resulting financial information and make it easier for a human to understand through natural-language explanations and summaries. Ollama provides the local runtime through which the application communicates with Gemma. This was an important part of the project because I wanted to understand how an application can be built around an open-weight model running locally, rather than treating AI as a black-box cloud API. 🔒 Local-first by design Financial information is sensitive. Since this project was built for my own family, I wanted privacy to be a meaningful part of the architecture. The AI inference happens locally through Ollama, and the financial data can remain on the user's machine rather than requiring a cloud AI service. This project is a great example of why I think open innovation matters, especially for applications dealing with personal data. If I had built FinTrack AI entirely around a closed AI API, the AI portion would essentially be: Send financial information → External API → Receive response Using an open-weight model changes that possibility. With Gemma 3 1B + Ollama, I can run the AI locally and control how the financial information is processed and provided to the model. That gives me the freedom to: Run AI locally Experiment with an open-weight model Control the context given to the model Decide how financial calculations and AI explanations interact Build without requiring a proprietary AI API Learn how local AI inference works in a real application For a financial application, this is especially meaningful because the data belongs to the user. Open-weight AI allowed me to explore a model that isn't just something I call through a remote endpoint — it can actually become part of the application running on the user's own machine. It also made this project a much better learning experience for me. I didn't just want to build an application that uses AI. I wanted to understand how to build an application around an open AI model. Best Use of Gemma Built by: Vaishnavi pophalkar ❤️ Why this project matters to me FinTrack AI started with a very simple problem at home. My father and uncle were already maintaining their financial records in Excel. I didn't want to tell them: "Stop using Excel and learn this completely new finance app." Instead, I thought: "What if I build something that works with the Excel sheets they already have?" That became the foundation of FinTrack AI. Now, instead of manually maintaining and analyzing everything, they can upload their existing data, understand it through dashboards and AI, ask questions naturally, and still get their data back as an Excel file whenever they need it. For me, that's what Build for a Friend is about. Not just building something technically impressive. Building something for a real person you care about, around a problem you've actually seen. ❤️