AI Commerce Assistant.
Turn product questions and screenshots into assisted orders.
← ALL PROJECTSAn e-commerce business received customer inquiries outside its team's working hours. Product questions and repetitive order handling also consumed time during the day.
I built an AI sales and support assistant for messaging channels. It answers product questions, retrieves product information, handles order actions and connects conversations to operational records. The client's identity is withheld.
- Contextual, multilingual product conversations
- Product recommendations, availability and images
- Product identification from customer screenshots
- Order creation, confirmation and cancellation
- Customer information capture and team notifications
- Messaging, database and business-system integrations
A customer sends a question, product URL or image through a connected messaging channel.
For visual lookup, product-video frames are sampled with OpenCV and embedded with CLIP. Vector similarity search retrieves candidates, with a vision-model check where needed.
The assistant returns relevant product information or handles an order action, updates the connected records and notifies the team.
TOOLS & TECHNOLOGIES Python · Async processing · OpenAI API · Supabase / PostgreSQL · pgvector · OpenCV · CLIP · Messaging APIs
- Use product retrieval to connect a customer's screenshot to catalog information.
- Keep conversations connected to operational records so order actions are represented outside the chat.
- Use messaging-channel adapters around the core assistant so conversation handling and product retrieval can share the same logic.
A completed client system connecting product discovery, customer support and order handling. Customer conversations lead to product answers and updates in the connected business records.
COMPLETED PROJECT · DEMO ASSETS TO FOLLOW