Inspiration
Ad teams often waste budget before realizing a creative is weak or fatigued. We wanted to build a copilot that helps marketers act earlier.
What it does
Smartex scores creatives by campaign KPI, detects fatigue/wearout, explains performance, and recommends practical creative changes such as cleaner layouts, stronger hooks, better CTAs, or new asset variants.
How we built it
We combined KPI-aware ML models, day-7 performance signals, fatigue detection, OCR/layout features, gaze-inspired analysis, explainability tools, and an LLM agent that turns model outputs into clear marketing recommendations.
Challenges we ran into
Avoiding leakage, handling misleading dataset variables, making recommendations realistic, and translating technical features into useful creative advice.
Accomplishments that we're proud of
We built an end-to-end creative intelligence system: scoring, fatigue monitoring, recommendations, explanations, and a dashboard/agent experience.
What we learned
Good ad intelligence is not just prediction. It needs context, uncertainty, explainability, and actionable recommendations.
What's next for Smartex
With real edit histories and A/B test feedback, Smartex could evolve into a contextual bandit/RL system for continuously improving creatives.
Built With
- langchain
- python
Log in or sign up for Devpost to join the conversation.