Invendora: AIを搭載したB2B検索プラットフォーム

MicrocosmWorks built Invendora's search using ChatGPT for conversational product queries combined with Elasticsearch for structured filtering and a vector database for semantic matching. Instead of keyword-based catalog browsing, buyers describe what they need in natural language and the AI returns contextually relevant products with explanations of why each matches. This approach surfaces products that keyword search would miss, especially for complex specification requirements.

MicrocosmWorks built a web scraping pipeline using Puppeteer, Cheerio, and Apify that crawls supplier websites, extracts product specifications, pricing, and availability data, and normalizes it into Invendora's unified schema. The pipeline runs on scheduled intervals with change detection to keep product data current. Extracted data is stored in PostgreSQL with vector embeddings indexed for semantic search, ensuring the catalog stays comprehensive and up-to-date.

Yes, MicrocosmWorks implemented a recommendation engine that analyzes purchase history, search patterns, and product affinities using the vector database. When a buyer searches for a product, the system cross-references their procurement profile to surface complementary products, preferred vendor alternatives, and quantity-based pricing tiers. LLM prompt engineering enables the AI to explain recommendations in the context of the buyer's specific use case.

MicrocosmWorks integrated SendGrid for automated email workflows between buyers and vendors. When a buyer expresses interest in a product, the platform generates a structured inquiry with specifications and quantity requirements, routes it to the appropriate vendor contact, and tracks the response pipeline. Vendors receive qualified leads with buyer context, while buyers get centralized communication tracking across all their product inquiries.

MicrocosmWorks built Invendora's core platform including the AI search engine, web scraping pipeline, vendor management system, and buyer-facing interface. At MicrocosmWorks development rates of $20-$40/hr, a similar AI-powered B2B search platform typically costs between $20,000-$45,000 for the initial build, with ongoing costs for scraping infrastructure, AI API usage, and platform maintenance available through our monthly support plans.