An AI-powered sales configurator for rule-compliant quotes

An internal web app for the sales team of a global cloud security (SASE) vendor. Sellers answer a guided questionnaire; the system recommends a bundle, calculates SKUs and quantities, and exports the quote.

Quotes had to respect a large set of product rules (incompatibilities, dependencies, min/max quantities, bundles, pricing and discounts), and admins needed to update those rules without a developer.

• Defined a monorepo architecture (shared / backend / frontend) running Nginx, NestJS, and Redis on Docker Compose, with S3 for files.
• Designed the core domain: questionnaire flow, recommendation engine, constraint validation, and pricing rules.
• Integrated an AI agent (Mastra with OpenAI GPT models) that recommends products from the answers.
• Built an admin import pipeline: spreadsheet parsers, a parser registry, and a BullMQ background queue.
• Secured access with Okta SSO and enforced test-coverage gates in CI.

• Sellers build rule-compliant quotes and export them as CSV or PDF.
• Admins keep the catalog and selling rules current by uploading spreadsheets.