Inspiration

I track my expenses in a low-code no-code tool called Quickbase and I thought there may be meaningful and interesting insights that an AI can derive as Gemini is well-versed in Data Analysis of structured data.

What it does

When dynamically queried expense data is fed to the Gemini model, it becomes capable of answering any questions to gain insights out of it. I created a chatbot to get such insights.

How we built it

Programming Language: Python Generative AI Model Used: gemini-1.0-pro Framework Used: Vertex AI SDK Web Service Framework: Google Cloud Functions Dataset Used: Personal Expenses tracked in Quickbase DB Frontend: Quickbase Dashboard

A web service is created and hosted with Google Cloud Functions making use of the standalone Python code initially created to test the functionalities. It is then used with the Bot interface created using HTML/JS and hosted as a web page in Quickbase. It is embedded in the Quickbase dashboard for quick access.

Challenges we ran into

Being the first time using the Gemini models and Cloud Functions, it took a while to understand them how to configure. But, Google documentation along with the Gemini search helped to accomplish them with little effort.

Accomplishments that we're proud of

Creating the Python code making use of the Gemini model, creating Cloud Functions to create a web service that can be used in Quickbase in a quick turnaround making use of Google Cloud services was an exhilarating experience

What we learned

The steps followed to create different Google Cloud services along with the first-hand experience of the Gen AI model to create a useful service were a great learning experience

What's next for Expense Analysis using Gemini

Creating a fine-tuned model to reduce the tokens that are consumed to use it while the chat conversation is made. This helped to understand how the Quickbase data be dynamically queried and analysed using Generative AI opens a lot of possibilities for many business applications created using the platform.

Built With

Share this project:

Updates