| Field | Details |
|---|---|
| Provider | MEMOIZE AI LLC, 5534 Saint Joe Road, Fort Wayne, IN 46835, USA |
| System Name | Memoize AI Memory-as-a-Service |
| Underlying Models | Google Gemini 2.5 Flash for response generation; Google text-embedding-004 for semantic memory embeddings (both via Google Vertex AI API) |
| Architecture | Retrieval-Augmented Generation (RAG) — the AI model retrieves contextually relevant memory data from vector databases to enhance response relevance |
| Capabilities | Personalized e-commerce product recommendations; contextual customer support responses; long-term shopper preference memory and recall |
| Limitations | The system may hallucinate or generate inaccurate information. Outputs should be verified before being relied upon for health, medical, or financial decisions. Not suitable for high-risk categorization. |
| Training Data | MEMOIZE AI LLC does not train the AI model. Customer and End User data is used solely for context retrieval (RAG) and never used to develop, train, or fine-tune any model. |
| Human Oversight | AI-generated outputs are recommendations only and are not reviewed by a human before delivery to End Users. Merchants and End Users retain full decision-making authority. Merchants may configure system prompts and product catalogs to shape AI behavior within their stores. |
| AI Governance | MEMOIZE AI LLC maintains an AI governance program including bias testing, output monitoring, and periodic model evaluation to ensure the Services operate within documented parameters. |
Input Data: When an End User sends a message through the chat widget, the system processes the message text along with contextual data (browsing context, conversation history, and relevant memories retrieved via vector similarity search).
Output Data: The AI generates a personalized text response that may include product recommendations, answers to questions, and acknowledgment of previously stored preferences or interactions.
Memory Storage: The system extracts key preferences, health constraints, and interaction patterns from conversations and stores them as 768-dimensional vector embeddings in a PostgreSQL database with pgvector. These memories persist across sessions and enable personalized future interactions.
Zero Data Retention: MEMOIZE AI LLC has enabled Zero Data Retention (ZDR) on Google Vertex AI. Prompts and responses sent to the AI model are not stored by Google after processing and are not used by Google to train or improve its models.
The AI system is designed exclusively for e-commerce personalization in the supplement and wellness space. The following uses are expressly prohibited:
Given the supplement and wellness focus, the system may process consumer health data including allergies, dietary preferences, and supplement regimens. This data is processed under the following safeguards:
For questions about this AI System Card or MEMOIZE AI LLC’s AI governance practices, contact: