Matheus Arrais
Worldwide Tech Leader, Cloud Operations & Principal Solutions Architect, AWS
Teaching cloud operations to run themselves, safely.
With 20 years in IT, I turn complex cloud ecosystems into resilient, self-healing environments. I lead AWS's worldwide Cloud Operations technical community — 2,200+ specialists designing operational solutions at scale — with deep expertise in AIOps, agentic operations, and modern observability.
I own the Cloud Operations track at AWS re:Invent, AWS Summits and AWS re:Inforce. I also run the AWS Cloud Operations Blog channel, which publishes 50+ posts a month from authors across AWS: 13,000+ posts to date and over 10 million unique views.
AgenticOps · AIOps · Observability
My current focus is autonomous agents for operations that enterprises can actually trust: pairing agentic incident response with the governance, guardrails, and audit trails that make autonomy safe in production.
I co-authored the AWS Certified Database Study Guide: Specialty (DBS-C01), and I write about agentic operations, cloud operations, observability, and DevOps.
What I'm working on.
Updated October 2026- Researching
PhD in Artificial Intelligence at the University of the Cumberlands, focused on AI for Operations and agentic operations.
- Speaking
AWS re:Invent 2026: one breakout session and one lightning talk.
- Leading
The AWS Cloud Operations technical community, and the Cloud Operations track at re:Invent, Summits and re:Inforce.
Agentic operations, in two halves.
AI for Operations
Agents that run operations, inside guardrails.
- 01 · Detect
Agents watch the signals and catch problems before customers do.
- 02 · Diagnose
They connect related alarms and find the likely cause.
- 03 · Act
Routine, reversible fixes run on their own: restart, scale out, roll back.
- 04 · Guardrails
Risky changes escalate to a person, and autonomy grows as trust is earned.
Operations for AI
Running AI workloads in production with the same rigor.
- 01 · Observe
Trace every model and agent call: latency, token usage, errors and cost.
- 02 · Evaluate
Measure answer quality and drift, not just uptime.
- 03 · Guard
Apply safety guardrails and keep an audit trail of what agents did.
- 04 · Optimize
Keep cost and capacity predictable as usage grows.
From the blog.
- Account-Level Retention Policies in CloudWatch Logs
CloudWatch Logs can now set retention for a whole account by prefix. Why a short policy deletes logs you meant to keep, and how to plan the set first.
- AWS DevOps Agent Now Takes Questions in Your Incident Channel
AWS DevOps Agent now takes questions in a private Slack channel. What that changes for on-call, why the channel now has an IAM role, and what to check before you turn it on.
- Your Traces Can Now Reach CloudWatch With One Header
CloudWatch now accepts API keys for OpenTelemetry traces. Where that removes real friction, why it's a long-lived credential, and what to set up before you hand one out.
Two decades, in four figures.
- In IT since2007Operations, cloud and security
- Community led2,200+Cloud Operations specialists, worldwide
- AWS Certified10×Professional · Specialty · Associate
- Published16Wiley book · my AWS blog posts, 25k+ views