TargetBoost AI

• 

  Final front-end on Google Sites

• 

  Google Form used to collect product descriptions

• 

  Google Sheet showing manual labeling

• 

  Colab example

• 

  Multi-Agent System Logic: Data Cleaner, Prediction, and Response Agents

• 

  TargetBoost AI System Architecture Overview

• 

  Gemini prompt & output

• 

  Product Description Input and Targeted Audience Output Interface

The idea behind TargetBoost AI was born from a real challenge faced by entrepreneurs in countries like Sudan those who are either banned from using online advertising platforms or simply can’t afford a marketing budget. As a beginner in digital marketing and AI, I joined this hackathon with the goal of learning and experimenting. I had no technical background, no laptop, and no prior experience — only a strong desire to solve a real-world problem. To build this project, I explored tools like Google Cloud, Python, Gemini, Google Forms, Google Sheets, and Google Sites entirely from my mobile phone. I started by collecting real product descriptions from local entrepreneurs using a Google Form. I then manually cleaned and labeled the data in Google Sheets, identifying the most relevant target audience for each product. This process became my training dataset. The backend was designed using Python on Google Cloud and structured around three simple agents: one for cleaning the data, one for generating predictions, and one for returning the final output. I attempted to train the model on Google Colab, but due to technical limitations — especially copy-paste errors and lack of hardware — I couldn’t complete the training process. As a workaround, I used Gemini AI to simulate the model’s logic and generate real-time targeting suggestions based on the labeled data. The front-end was built using Google Sites, allowing users to input their product description and instantly receive AI-powered audience recommendations without needing any budget or technical skills. This project taught me that passion and persistence matter more than tools. Even without a laptop, funding, or prior experience, it’s possible to build a meaningful AI solution. Note: The current Google Sites page serves as a presentation-only interface to demonstrate the system’s logic and workflow. Due to technical constraints (no access to full hosting or desktop tools), live input/output functionality is provided through the linked Google Form and Gemini response examples. The system is designed to evolve into a fully interactive mobile or web application in future versions. And even if I don’t win, I’m proud that I completed this project against all odds. This journey taught me that self-learning, persistence, and a sincere intention to solve a real problem can lead to meaningful impact even with no resources. I'm also open to feedback, collaboration, or internship opportunities to grow this project beyond the hackathon.

Built With

• cloud
• forms
• google
• google-cloud-(vertex-ai
• no-code
• sheets
• storage)