Price: 0
Number of applications: 7
27.02.26 (inclusive)
Monetary
Finished product
ICT tasks
Light industry
Neurotechnology and artificial Intelligence
Software/ IS
The current search on the customer's website (an e-commerce platform for finishing materials) is limited by strict keyword matching and does not take into account the visual preferences of customers. Users often: 1. Do not know the exact name of the material ("loft-style tile"). 2. They have only a photo reference without characteristics. 3. They need advice on the compatibility of materials, which requires the participation of an expensive consultant. This leads to a low conversion rate from the search to the shopping cart and a limited average receipt due to the lack of cross-sales.
Conversion Rate (CR) growth: An expected 15-20% increase due to a more accurate hit in the customer's query. Increase in the average receipt (AOV): Due to AI articles that offer comprehensive solutions (for example, suitable grout and glue for tiles). Reducing the support burden: Automating basic style selection consultations.
Agafonova Julia
Purpose and description of task (project)
Goal: To develop an AI-based service (LLM + Vision models) for implementing multimodal search functionality and personalized recommendations on the e-commerce finishing materials platform. Description: The system must accept text queries or images as input, conduct a semantic search through the customer's product catalog and generate an expert advice article in real time. The delivery is carried out as an isolated Docker container for deployment in the customer's infrastructure (on-premise).
Whatsapp
