Loading post…
GrabSneh AgrawalProblem & solution
From firefighting to building: How AI agents restored our team’s core productivity
Summary
Grab’s Analytics Data Warehouse team supports more than 1,000 monthly users and 15,000 tables, but repetitive questions about definitions, lineage, quality, and basic enhancements consumed about 40% of engineering time. It deployed a multi-agent AI system that routes Slack requests into enhancement or investigation pathways, using FastAPI and LangGraph for stateful, cyclical collaboration, Redis for caching and sessions, and PostgreSQL for persistent memory. Specialized agents query Trino, Hive, or Delta Lake, inspect GitLab code, check incidents and pipeline health, or prepare semi-automated code changes, while human reviewers approve production modifications. The system autonomously resolves most standard inquiries and some common enhancements, reducing resolution time by an order of magnitude and bringing simple answers to users within minutes. The resulting capacity shift reclaimed several FTEs worth of engineering bandwidth and moved hundreds of hours from reactive support to roadmap work.
Context
The Analytics Data Warehouse team was spending approximately 40% of its time handling repetitive requests about data definitions, lineage, troubleshooting, quality checks, and basic enhancements. Investigating seemingly simple questions required navigating catalogs, tracing code and lineage, validating SQL, and reviewing logs, creating a support bottleneck.
Approach / What changed
A multi-agent AI system receives requests through Slack and routes them to either an Enhancement Agent or an investigation workflow. LangGraph coordinates specialized agents, while Redis and PostgreSQL manage session and persistent state. Agents use data, code, metadata, quality, observability, and collaboration tools; engineers retain final approval over production code changes.
Takeaways
The system separates enhancement requests from investigation questions, then coordinates specialist agents instead of relying on one monolithic AI.
Enhancement work remains semi-automated: the agent gathers context, proposes code, creates a merge request, runs tests, and leaves review and merge decisions to engineers.
The deployment reduced issue-resolution time by an order of magnitude, resolved simple inquiries within minutes, and reclaimed several FTEs worth of engineering bandwidth.
Source and freshness
- Source
- engineering.grab.com
- Published
- Added to Yomu
- Last updated
Summaries are generated from the original by a language model. Check the original before relying on it.
Keep reading
Related reading
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
The best engineering reads, once a week
One useful email. Unsubscribe anytime.