MemCommerce AI

Inspiration

What it does

How we built it

Challenges we ran into

Inspiration

Modern e-commerce platforms often struggle to deliver personalized shopping experiences because standard AI chatbots lose context between sessions. Customers have to repeat their preferences, order inquiries, and styling choices every time they interact with the shop. We wanted to build an autonomous e-commerce AI assistant that retains persistent "agentic memory" across sessions—remembering user preferences, inventory updates, and interaction history in real-time.

What it does

MemCommerce AI is an intelligent, memory-enabled assistant for online retail. Powered by persistent agentic memory, it:

• Tracks User Preferences: Remembers customer styling preferences, preferred sizing, and past queries without relying on temporary session storage.
• Syncs Real-Time Inventory: Connects directly with product databases to give live stock updates, product recommendations, and context-aware responses.
• Delivers Context-Aware Support: Uses past interaction logs to proactively assist buyers with order tracking and tailored recommendations.

How we built it

• Database & Memory Engine: Built on CockroachDB Serverless to store high-performance vector embeddings, persistent user context, and transactional product inventory with strong consistency.
• Cloud & Hosting: Hosted on AWS services for scalable, high-availability serverless deployment.
• AI & LLM Orchestration: Integrated LLM frameworks with custom agentic memory workflows, enabling the agent to perform long-term context storage and instant retrieval.

Challenges we faced

Structuring state management for long-running agentic conversations was a key hurdle. Standard LLM setups suffer from context-window limits. We had to design an efficient data schema in CockroachDB to index, retrieve, and score conversation history dynamically so the AI agent gets exact relevant memory without hitting context limits.

Accomplishments that we're proud of

• Successfully implemented real-time persistent agentic memory that survives across separate browser sessions.
• Seamlessly combined relational inventory state with vector-based conversational context using CockroachDB.
• Created an intuitive, ultra-fast UI experience for online shoppers.

What we learned

We gained deep insights into designing resilient distributed agent architectures using CockroachDB and AWS. Managing state for AI agents requires balancing relational accuracy (for product data/orders) with unstructured context memory (for conversational history).

What's next for MemCommerce AI

• Multi-Platform Integration: Expanding connectors to sync directly with platforms like TikTok Shop and seller centers.
• Proactive Engagement: Enabling the AI to reach out autonomously with personalized offers based on restocked items in user wishlist histories. ## Accomplishments that we're proud of

What we learned

What's next for MemCommerce AI

Built With

• amazon-web-services
• cockroachdb
• llm
• node.js
• python
• react
• sql
• tailwinds-css
• vector-database