SuperAds

I built a custom AI-powered advertising engine that lives entirely within our platform. Instead of relying on third-party ad networks

Comment

Inspiration

We noticed that most ad systems rely heavily on third-party networks, which often lack personalization and control. We wanted to create a system that delivers ads directly on the platform, leveraging AI and machine learning to serve relevant content based on a user’s profile and interests.

What it does

SuperAds is a custom ad recommendation engine built into the website. It uses AI-generated ads descriptions and an XGBoost machine learning model to categorize ads, then matches them to users based on their profile. This ensures every ad is relevant, personalized, and fully controlled by the platform.

How we built it

Backend: FastAPI for serving ad recommendations via APIs Database: PostgreSQL for storing structured user and ad data Cache: Redis for low-latency access to frequent user profile information Machine Learning: XGBoost model to classify ad categories based on AI-enhanced descriptions Workflow: Users set up their profile → AI processes ad metadata → XGBoost predicts category → system recommends ads

Challenges we ran into

Ensuring real-time recommendations without slowing the user experience Handling string and categorical features in XGBoost efficiently Overcome LLM lack of structure of response utilizing LLM like OpenAI

Accomplishments that we're proud of

Successfully combined AI content generation with machine learning category prediction

What we learned

How to integrate AI and ML into a production system or API

What's next for SuperAds

Introduce A/B testing to measure recommendation effectiveness Add a dashboard for analytics, showing ad engagement and category performance Take pass orders into account, and how often users click on ads, and modify the model accordingly

Built With

Share this project:

Updates