IcarIA

Campaign Analysis Platform_definitions
Campaign Analysis Platform_results
Campaign Analysis Platform

Inspiration

We realized that while AdTech generates mountains of data, designing new creatives remains a guessing game. Our inspiration was to eliminate this uncertainty. By building a prescriptive model that decodes which specific design elements maximize KPIs and predicts ad fatigue, we empower Smadex advertisers to engineer highly optimized, data-driven campaigns from day one.

What it does

IcarIA is a multimodal auditing engine that automates quality control for digital ads. It simultaneously sees images and reads ad copies to ensure they align with the specific business category. If the AI detects a discrepancy in the business objective, it flags the inconsistency, preventing dirty data from corrupting performance prediction models.

How we built it

We built the core logic using Python and Streamlit for a seamless UI. The motor of the project is Gemini, which we leveraged for its multimodal capabilities to process text and images in a single pass. Then we used GXBoost to evaluate the ads and generate the best approach a company can have.

Challenges we ran into

The biggest problem was understanding the datasets the company provided us. As we learned more and more about it, we found out that there was a lot of dirty data, making it difficult to train our model.

Accomplishments that we're proud of

We successfully transformed a messy, subjective manual task into a structured, automated pipeline. We are particularly proud of our Inconsistency Detector, which doesn't just extract data but actually understands when a creative "feels wrong" for a specific vertical.

What we learned

We learned that in Data Science, the "Garbage In, Garbage Out" rule is king. No matter how good your prediction model is, it’s useless without clean data. We also mastered the art of multimodal prompting—learning how to guide an AI to analyze visual and textual context simultaneously to reach a logical conclusion.

What's next for IcarIA

Having mastered the auditing and correction phase, the next step is to expand into Creative Generation. We aim to implement a feature that generates detailed design briefs and optimal ad copies based on successful patterns. Furthermore, we plan to integrate Image Generation to create missing assets and a Video Processor to audit and optimize video ads with the same level of precision we currently apply to static images.