La Cómplice, a tech startup, aimed to revolutionize how e-commerce interacts with influencer-driven content. They envisioned an AI powered ‘magic’ solution that could identify specific fashion items from social media photos and match them with inventory from fashion stores. Conversely, they wanted to show how a product might look when styled by influencers.

They work with some of the best clothing brands in the world, so the challenge was huge.

Tailor Hub delivered a cutting-edge image-matching AI tool capable of:

• Bidirectional matching, extracting item features from influencer photos and matching them with their clothing inventory databases. Then, the tool is capable of mapping those items back onto influencer images.

• Advanced image segmentation: tool was able to identify individual elements (bags, dresses...) from composite photos where models wear multiple items.

• Human-in-the-loop: we developed a manual review tool that allows our client to fine-tune the tool and matching system for greater accuracy.

We used GPT-4 and GPT Vision, Chroma for vector database management and LangChain to orchestrate data and models..

The solution achieved remarkable precision and enabled us to keep expanding new AI features for the biggest retailers in the world.