From Paper to JSON - GDG Bucharest

| 🧑🏻‍💻 Code | 🖥️ Presentation

Earlier this month, I had the chance to present at the Google Developer Meetup in Bucharest, alongside my longstanding friend and mentor Lucian Gruia, where I introduced a workflow for digitizing and processing expense receipts using AI.

In the end the task deemed to be fairly straightforward and simple but the presentation mainly aimed highlighting how easy is to introduce AI powered workflows in our apps and in our day to day activities.

The Problem

Being part of a big friend group, closing trips or other group activities proved challenging when it came to cost splitting. We were always reliant on expense-splitting apps (e.g., SplitWise). Lately, I've taken on the challenge of building such an app and discovered just how complex, yet rewarding, it is to create an automated solution that not only splits expenses but also itemizes receipts, tracks debts, and handles multiple currencies. The process opened my eyes to how much we can leverage AI and modern backend systems to make group activities smoother and more enjoyable.

The Solution: From Paper to JSON

My presentation demonstrated how to digitize these paper receipts and convert them into structured JSON, ready for backend processing. Here's how it works:

1. Scan Receipts: Start with scanning the physical receipts.
2. LLM-Powered Itemization: Use Large Language Models via Vertex AI to extract and itemize data from receipts.
3. Generate JSON: Convert the extracted data into structured JSON.
4. Backend Integration: Feed the JSON into backend systems for processing and validation.

Results

This workflow significantly enhances the user experience by simplifying complex, time-consuming processes into quick and seamless tasks. For the everyday user, this means no more struggling with manual data entry or dealing with frequent mistakes that slow them down.

By automating tedious processes, the workflow empowers users to focus on decision-making and more meaningful actions rather than getting bogged down in repetitive tasks. The result is a streamlined experience that feels intuitive and supportive, elevating the overall satisfaction and ease of use for anyone interacting with the system.

Presenting this solution at the GDG Bucharest meetup was an incredible experience, and I'm excited to see how AI continues to transform data processing workflows.

| 🧑🏻‍💻 Code | 🖥️ Presentation