3DFanatics: plataforma inteligente de e-commerce para emprendedores de impresión 3D en Colombia

In Colombia, small and medium-sized 3D printing companies face low visibility, high fees, and fraud risks when operating on generalist e-commerce platforms. This article introduces 3DFanatics, an intelligent platform prototype designed to optimize user experience and enhance security through personalized recommendations, automated conversational support, and biometric verification. The agile Scrum methodology was adopted across eight sprints to rapidly iterate on design and implementation. The architecture was deployed on Amazon Web Services using the Cloud Development Kit (CDK): AWS Cognito managed identities; API Gateway and Lambda functions handled business logic; RDS and S3 stored data; and the React JS frontend was hosted on Vercel. The recommendation layer was implemented with AWS Personalize, trained on a synthetic dataset of 25 users, 16 products, and 1,000–10,000 interactions. This system was evaluated using normalized discounted cumulative gain (NDCG), precision, mean reciprocal rank (MRR), and coverage metrics, achieving a precision of 0,75 when generating five suggestions. The chatbot, developed with AWS Lex, supported CRUD operations for inventory and orders, though it exhibited limitations with unanticipated queries. Identity verification integrated AWS Rekognition with an 80 % liveness threshold and a 75 % facial-comparison threshold. Additionally, AWS Textract extracted data from national ID cards with an 80 % confidence threshold. The validation process included unit tests with Jest and Pytest, integration tests, end-to-end tests with Cypress, and think-aloud usability sessions with five industry entrepreneurs. A cost analysis for the development period (October 2024–March 2025) revealed total expenditures below USD 50, demonstrating the prototype’s economic feasibility. In conclusion, 3DFanatics validates its technical and economic viability as a specialized 3D-printing e-commerce solution, and further evaluation of its scalability and robustness with real-world data in production environments is recommended.