Incident management · Production

Grab AI Gateway: Connecting Grabbers to multiple GenAI providers

The problem

Before Grab AI Gateway, teams faced fragmented AI provider access with incompatible authentication schemes, wastage from siloed reserved capacity, no central compliance auditing, and no unified cost attribution across use cases.

Workflow diagram · grounded in source

Request exploration key via Slack

trigger

“These keys can be requested by Grabbers through a Slack bot.”

Mini-RFC onboarding review

human_review

“Every new use case requires a mini-RFC (Request For Comments) and a checklist that is reviewed by the platform team. In certain cases, an in-depth review by the AI Governance task force may be requested.”

Authentication and authorisation

validation

“The gateway handles functionalities like authentication, authorisation, and rate limiting, allowing users to solely focus on building GenAI enabled applications.”

Dynamic routing to providers

routing

“It provides the control points to dynamically route requests for certain models to a different albeit similar model backed by a reserved instance. Another frequent use case is smart load balancing across different regions to address regi…”

Cost attribution and audit logging

output

“The AI Gateway records each call's request, response body, and additional metadata like token usage, URL path, and model name into Grab's data lake.”

Reported outcome

Grab AI Gateway has onboarded more than 300 unique use cases to production, with more than 3,000 Grabbers requesting exploration keys, powering applications including real time audio signal analysis, content moderation, translation, and incident management automation.

Reported metrics

Unique use cases onboarded to productionmore than 300

Grabbers requesting exploration keysmore than 3,000

Reported stack

OpenAIAzureAWSBedrockGoogleVertexAISlackChimera notebooksCatwalkVLLM

Source

https://engineering.grab.com/grab-ai-gateway

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Grab AI Gateway has onboarded more than 300 unique use cases to production, with more than 3,000 Grabbers requesting exploration keys, powering applications including real time audio signal analysis, content moderatio…

What tools did this team use?

OpenAI, Azure, AWS, Bedrock, Google, VertexAI, Slack, Chimera notebooks, Catwalk, VLLM.

What results were reported?

Unique use cases onboarded to production: more than 300; Grabbers requesting exploration keys: more than 3,000 (source-reported, not independently verified).

How is this incident management AI workflow structured?

Request exploration key via Slack → Mini-RFC onboarding review → Authentication and authorisation → Dynamic routing to providers → Cost attribution and audit logging.