Inspiration
Digital marketing teams and e-commerce store owners spend hours brainstorming unique social media ad copies for huge lists of products. This manual work creates a massive slowdown when trying to launch quick marketing campaigns. We were inspired to build a tool that automates this entire pipeline safely, securely, and instantly.
What it does
AdFlow is a full-stack marketing campaign generator. A user simply types in a product name, and the system instantly writes tailored social media marketing copy, headlines, and optimized hashtags. While displaying the result to the user, the application simultaneously handles cloud file replication and database history logging under the hood.
How we built it
We built AdFlow using a clean, production-grade tech stack:
- Frontend UI: Engineered a responsive web browser application using Streamlit.
- AI Intelligence Layer: Integrated Google's Gemini 2.5 Flash model via the Google GenAI SDK to draft high-converting ad layouts.
- Cloud Infrastructure: Integrated Boto3 to instantly mirror and back up copy configurations into a secure Backblaze B2 cloud object storage bucket.
- Persistence Layer: Embedded a local SQLite database to automatically keep track of past campaigns with custom data indices.
Challenges we ran into
This being a deep pipeline project, we ran into realistic backend challenges. We had to manage deprecated API modules and cleanly refactor our logic to stick to Google's newest GenAI SDK endpoints. Additionally, we handled layout loop states inside our web application threads when synchronizing database transactions, which we resolved by optimizing Streamlit’s state reloading mechanics.
Accomplishments that we're proud of
We are proud of building a fully functional, end-to-end full-stack pipeline that connects local operations with enterprise cloud systems seamlessly. Every file successfully pushes to cloud buckets while indexing into our local structural database instantly without crashing the interface.
What we learned
We learned how to design a secure software architecture by decoupling raw secret strings into separate .env configuration files to prevent credential leaks. We also mastered systematic debugging across different API protocols, handling boto3 cloud parameters, and managing relational database schemas using SQLite.
What's next for AdFlow
Next, we want to upgrade the prompt intelligence layer to support proven marketing frameworks like AIDA (Attention, Interest, Desire, Action). We also plan to expand the interface to generate matching marketing graphics alongside the text copy using multi-modal models.
Log in or sign up for Devpost to join the conversation.