Marketeer

An AI-powered Marketing Platform that leverages A/B testing to optimize email campaigns and ensure they reach their full potential.

Comment

✨ Inspiration

Email marketing often fails due to overloaded inboxes, causing newsletters to be ignored and reducing their overall impact. This project addresses that challenge by aiming to boost sales and increase click-through rates (CTR) through AI-crafted content.

🚀 What it does

This project introduces Marketeer, an AI-powered marketing platform.

By simply describing an event and adding it to the campaign calendar, Marketeer automatically generates an email campaign. Using Google's gemini-1.5-pro-001 model, it creates two unique email variants (A & B) for each event. Afterwards customers (from the Firestore users collection) are split into A/B test groups, and emails are sent out autonomously.

Each email includes a CTA button linking to a mock e-commerce store, designed to simulate a real environment and track user behavior such as opens, clicks, and purchases. These interactions are then visualized in real time on the Marketeer dashboard to help evaluate campaign effectiveness and optimize future outreach.

🛠️ How it is built

With Angular a frontend is created for the Marketing platform and the mock webstore. While Python is used and deployed as GCloud Run function for handling Firestore database operations, campaign generation, mail delivery, tracking, and analytics.

🧠 Challenges & Lessons Learned

One of the biggest challenges was the initial setup of the Google Cloud Run backend, especially since I originally planned to use AWS. However, because Gemini is part of the Google ecosystem, I made the switch and learned how to integrate services seamlessly.

Another hurdle was configuring the GenAI API properly. I discovered that only specific Gemini models were supported.

In addition, handling CORS (Cross-Origin Resource Sharing) issues proved to be particularly tricky during development and deployment. Ensuring that the frontend (hosted via GitHub Pages) could securely communicate with the backend APIs deployed on Google Cloud Run required fine-tuning response headers and endpoint permissions.

🏆 Accomplishments that I am proud of

  • Successfully deploying a full-stack system, from frontend and backend to database and GenAI integration, gaining experience in a Cloud First approach
  • Creating a workflow that autonomously generates and sends visually appealing marketing emails
  • Real-time analytics and behavior tracking for continuous optimization

🔮 What's next for Marketeer

  • Smart Email Scheduling: Currently, emails are sent immediately for demo purposes. The next step is to analyze user behavior (e.g., when they typically open emails) to deliver emails at the optimal time, ensuring visibility.
  • Behavior-Driven Personalization: Future improvements include enriching emails with social media trends, purchase history, and user behavior to deliver hyper-personalized content.
  • Fine-tuning & RAG: Customizing Gemini responses further with domain-specific knowledge about the store via Retrieval-Augmented Generation (RAG) or fine-tuning.
  • Prompt Engineering Enhancements: By refining the email generation prompt using few-shot prompting, including examples of the most successful past campaigns, the model can be guided to craft more effective content.
  • Security: Adding protection mechanisms against prompt injections
  • Time-Series Visualizations of metrics for better insight into performance trends.

Built With

Share this project:

Updates