Blime AI

Inspiration

Local vendors in Auchi and surrounding areas have incredible inventory from clothes and shoes to everyday gadgets but face a massive bottleneck when trying to sell online: the manual labor of digitization. Small business owners simply do not have the time or technical expertise to sit down and write professional, high-converting product descriptions for hundreds of items. We realized that for a local marketplace to truly scale, the onboarding process couldn't just be "easier" it had to be entirely automated. We needed a platform that ran its own catalog operations.

What it does

Blime AI is an e-commerce platform where the core operations are managed by "Xed," an autonomous AI agent. Instead of filling out complex product forms, a vendor simply takes a raw photo of their item and uploads it. Xed takes over immediately. Utilizing the Gemini API, Xed analyzes the image, identifies the product, and instantly generates a standard commercial item description. Xed then automatically categorizes the product (e.g., Gadgets, Clothing, Footwear) and pushes the listing live. Vendors can launch a fully professional digital storefront with zero manual data entry.

How we built it

To move fast and focus on the AI-native layer, we utilized our pre-existing PHP and SQL marketplace architecture as a foundational boilerplate for routing and database management. The entire AI operational layer was built fresh for this hackathon. We integrated the Gemini 1.5 Flash model via a PHP backend to handle multimodal inputs. When an image is passed to the server, it is sent to the Gemini REST API alongside a strict prompt engineering framework that forces Xed to return the analysis as structured JSON. Our backend parses this JSON and autonomously executes the database insertion, creating the live product listing.

Challenges we ran into

Getting consistent, professional output from the AI was our biggest hurdle. Initially, the generated descriptions were sometimes formatted incorrectly or included unnecessary conversational text. We solved this by refining our prompt engineering to strictly demand standard commercial descriptions and locking the API's output format to JSON. This ensured our PHP backend could reliably parse the data and insert it into the SQL database without breaking the application.

Accomplishments that we're proud of

We successfully transformed a traditional, manual e-commerce workflow into an AI-native operation. By handing the catalog management over to Xed, we have reduced the time it takes a vendor to list a product from several minutes down to just a few seconds. We are proud to have built an AI integration that executes real business tasks invisibly in the background

What we learned

We learned the distinct difference between an "AI-assisted" platform and an "AI-native" platform. AI-native means the artificial intelligence is making actual operational decisions and executing them autonomously. It forced us to rethink our entire database structure to allow an API, rather than a human, to be the primary data entry operator.

What's next for Blime AI

Catalog management is just Xed's first job. The next phase of Blime AI will integrate Xed into supply chain forecasting and dynamic matchmaking. Xed will soon monitor search trends across the platform, identify when a vendor has slow-moving inventory that matches those trends, and autonomously propose targeted flash sales to clear the stock acting as an automated sales manager for every local vendor on the platform.

HOW TO TEST THE AI (FOR JUDGES) Xed adapts based on the user's role. Please use these test accounts to experience the full AI: To test the Vendor AI (Image Upload & Stock, other features included): Login: [email protected] Password: password123 To test the Buyer AI (Search & Add to Cart, other features included): Login: [email protected] Password: password123