TrackMySpend is a web app that helps young Indian families make sense of their finances by turning raw bank and credit card statements into clear, categorized insights.

Users can securely upload their statements, which are parsed and structured using LLM-powered transaction processing. The app automatically cleans vendor names, detects duplicates, and categorizes expenses into meaningful groups with options for manual edits. A responsive dashboard then surfaces charts, summaries, and AI-generated insights that highlight spending trends and opportunities to save.

Privacy and trust were key considerations, with all processing built on Supabase, Vercel, and secure API integrations. The frontend was built with Next.js, Tailwind, and Shadcn, while Gemini and Claude handled parsing, vendor deanonymization, and insights.

I owned the full end-to-end build: setting up Supabase authentication and database, implementing the PDF parsing and categorization pipeline, building the editable transaction table, and creating the responsive dashboard experience.

Tech stack: Next.js 14, Supabase, Vercel, Tailwind CSS, Shadcn/ui, Gemini API, Claude, OpenAI.