Vertex AI Feature Store with ADK

Vertex AI Feature Store is a managed repository for ML features that enables low-latency, real-time feature serving. When combined with ADK agents, it allows agents to retrieve precomputed customer and product features - such as purchase history, loyalty tier, and browsing preferences - to personalise responses. ShopMax India uses this to power its recommendation agent.

The integration works by exposing Feature Store lookups as ADK tools. The agent calls these tools with a customer or product ID, retrieves the relevant features, and uses them to tailor its response or recommendation logic.

The below example shows how to create a Feature Store online serving client and register it as an ADK tool.

It gives the following output,

Feature Store connected: shopmax_customers
Entity type: customer_profile
Agent initialized: shopmax_recommender

The below example shows the agent retrieving customer features and generating a personalised product recommendation.

It gives the following output,

Customer CUST-BLR-00421 profile:
  Loyalty tier: Gold | City: Bangalore
  Avg order: Rs 18,500 | Preferred: Premium Electronics

Recommendations:
1. Sony 65 inch OLED TV - Rs 1,45,000 (matches premium electronics preference)
2. Bose QuietComfort 45 Headphones - Rs 28,000 (Gold tier exclusive offer)
3. Samsung Galaxy Tab S9 - Rs 62,000 (top seller in Bangalore this week)

ShopMax India serves personalised recommendations to over 50,000 daily active customers. By connecting ADK agents to Vertex AI Feature Store, the recommendation agent retrieves each customer profile in under 10ms and generates contextually relevant suggestions - increasing average order value by 22% compared to generic recommendations.