1 of 10

Google Cloud x MLB�Personalized fan highlights

My MLB powered by Vertex AI Gemini 2.0

Nishidh Singh Shekhawat

2 of 10

1a. User stories

Gemini can do all this and more!

Fan 1: Staying up to date on my games, Quick Recaps for My Busy Life

As a MLB fan, I want to receive personalised Highlights when i missed watching a game, so that I am always up to date while discussing with my MLB friends.

Fan 2: Game Day Recaps in My Native Language, News Just for Me

As an international soccer fan, I want personalized highlights in my preferred language and tailored to me, so I can quickly catch up on the match I missed

3 of 10

1b. Functionality based on User Journeys

Subscribe and Receive Digests:Ability to select favourite players and teams .

“Personalisation” of Digests: Every individual fan is different, tailored content for each persona

MLB Rule AI Search: Allows new fans to search for MLB Rules

Multi Language Support: Should support English, Japanese, Spanish.

4 of 10

2. Demo

My MLB powered by Vertex AI Gemini 2.0

https://mlb-frontend-service-513391239750.asia-south1.run.app/

5 of 10

3a. Designing the Google Cloud Architecture

Microservice Architecture:

Service 1: Handles all MLB Fan Interactions

Service 2: Handles all Gen AI Requests

Service 3: Sending Gmail Digests

6 of 10

3b. Designing the Google Cloud Architecture

Selecting the “Right” tools and Google Cloud services.

Compute:

-Cloud Run Service: Flexible and Serverless

-Cloud Run Functions: Perfect to trigger from Pub/Sub for real time MLB events

2nd Gen Cloud Run has lot more flexibility.

AI:

-Vertex AI: Many Models and Client libraries available

-Gemini 2.0: Personalized digests and is very fast.

-Agent Builder: No code to build MLB Rule Search

One of the first things to keep in mind is the cloud architecture. This has always helped me design reliable and scalable solutions.

We can divide the google cloud services into these 4 Sections.

Compute:

Cloud Run and Cloud Run functions were chosen as they are Serverless Compute services

While GKE offers Flexibility and App engine offers fully managed service,It takes more time to set up the GKE pods and Ingress as compared to ap engine.

Cloud run and cloud functions fit in perfectly for our use case of microservices

AI:

Google Cloud Vertex AI is the best choice, Many models in model garden , direct api availability and access via Client Libraries

We have used Gemini 2.0 to summarize, personalise and create the digest

Vertex AI agent builder is a really good no code platform,

with 5 minutes of Configuration and uploading the documents, i had a Agent who can answer any questions about any MLB rules

Database and Storage

For Database Cloud SQL was selected as it is Perfect Managed SQL for storing user metadata , like his preferred teams, players,language, who he/she as subscribed to.

For Storing Videos/images, digests for the users we have selected GCS since it is a low cost Object Storage perfect for our use case.

Network and Security:

Google OAuth 2.0 for Login to the Frontend Service

Secret Manager to store any credentials which will be accessed by the application.

Google has variety if Interconnects and VPC’s which can be used. For this use case since we have a simple Cloud Project Default VPC with configure firewall rules is being used.

For connecting our Privare Cloud SQL instance with our Frontend Service we are using Serverless VPC

Cloud Armor can be used with Cloud run to protect from DDoS Attacks.

7 of 10

3b. Designing the Google Cloud Architecture

Selecting the “Right” tools and Google Cloud services:

Database and Storage:

-Cloud SQL: Flexible and Serverless

-Cloud Storage: Store Images videos and digests

Network and Security:

-Google OAuth 2.0 SSO Login

-Secret Manager: Credentials, secrets storage

-Serverless VPC: Connect to Private Cloud SQL

-Cloud Armor: DDoS and ML based protections

One of the first things to keep in mind is the cloud architecture. This has always helped me design reliable and scalable solutions.

We can divide the google cloud services into these 4 Sections.

Compute:

Cloud Run and Cloud Run functions were chosen as they are Serverless Compute services

While GKE offers Flexibility and App engine offers fully managed service,

Cloud run and cloud functions fit in perfectly for our use case of microservices

AI:

Google Cloud Vertex AI is the best choice, Many models in model garden , direct api availability and access via Client Libraries

We have used Gemini 2.0 to summarize, personalise and create the digest

Vertex AI agent builder is a really good no code platform,

with 5 minutes of Configuration and uploading the documents, i had a Agent who can answer any questions about any MLB rules

Database and Storage

For Database Cloud SQL was selected as it is Perfect Managed SQL for storing user metadata , like his preferred teams, players,language, who he/she as subscribed to.

For Storing Videos/images, digests for the users we have selected GCS since it is a low cost Object Storage perfect for our use case.

Network and Security:

Google OAuth 2.0 for Login to the Frontend Service

Secret Manager to store any credentials which will be accessed by the application.

Google has variety if Interconnects and VPC’s which can be used. For this use case since we have a simple Cloud Project Default VPC with configure firewall rules is being used.

For connecting our Private Cloud SQL instance with our Frontend Service we are using Serverless VPC

Cloud Armor can be used with Cloud run to protect from DDoS Attacks.

8 of 10

4a. What’s Next: Imagen and VEO

VEO 2 in Vertex AI is in preview as of this date.

Imagen 3 and VEO 2 can further personalise the Image in the digest tailored to the user

9 of 10

4b. Summary

Fans receive personalised digests for their players, teams and in their native languages (English, Spanish, Japanese) using Gemini 2.0

MLB data sources have been effectively integrated to the services.

Vertex AI Search Allows new fans to search for MLB Rules

Google Cloud Services are deployed in a reliable and scalable microservice architecture.

10 of 10

Hey MLB Fan!

Scan the QR code

or

click the link to try My MLB

https://mlb-frontend-service-513391239750.asia-south1.run.app/

Google Cloud x MLB�Personalized fan highlights

My MLB powered by Vertex AI Gemini 2.0

Nishidh Singh Shekhawat