Quality assurance · Production

Atlassian builds Firecracker microVM platform in four weeks using Rovo Dev AI agents

The problem

Building a Firecracker microVM orchestrator on Kubernetes required constructing a scheduler, autoscaler, node agents, envoy ingress layers, raft persistence, and more — a scope previously considered too long, too complex, and beyond the team's domain expertise.

Workflow diagram · grounded in source
1
Multi-workspace agent setup
trigger
“Three workspaces, each checked out on a different branch, each with an agent working”
2
Agent writes code and e2e tests
ai_action
“AI writes the e2e tests too. The agent writes tests, deploys to a dev shard, runs them, and loops on failures until they pass. The test suite is the primary proof that things work.”
3
Dev shard deployment loop
validation
“Every feature gets deployed to an isolated dev shard on a real cluster. The agent deploys, tests e2e, fixes issues, redeploys.”
4
CI pipeline quality gate
validation
“Every PR runs lint, vet, tests, and Helm validation. The agent reads pipeline output and addresses failures before requesting review.”
5
Adversarial subagent review
ai_action
“For review, have an adversarial persona subagent that spins up and reviews what the main agent has written.”
6
Progressive production rollout
output
“Progressive rollout: main deploys to dev without PRGB, so we can validate internally fast. Production gets canary deploys across multiple clusters.”
Reported outcome

Atlassian built the Fireworks platform — a Firecracker-based microVM system with 100ms warm starts and live migration — in four weeks, entirely by LLMs using Rovo Dev agents.

Reported metrics
VM warm start time100ms warm starts
Platform build timefour weeks
code authored by AIentirely by LLMs
Scope expansion vs prior capabilitybuilding things we never would have committed to before
Reported stack
Rovo DevFirecrackerKubernetesBitbucketPipelinesHelmeBPFAWS
Source
https://www.atlassian.com/blog/rovo/rovo-dev-platform-driven-development
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Atlassian built the Fireworks platform — a Firecracker-based microVM system with 100ms warm starts and live migration — in four weeks, entirely by LLMs using Rovo Dev agents.

What tools did this team use?

Rovo Dev, Firecracker, Kubernetes, Bitbucket, Pipelines, Helm, eBPF, AWS.

What results were reported?

VM warm start time: 100ms warm starts; Platform build time: four weeks; code authored by AI: entirely by LLMs; Scope expansion vs prior capability: building things we never would have committed to before (source-reported, not independently verified).

How is this quality assurance AI workflow structured?

Multi-workspace agent setup → Agent writes code and e2e tests → Dev shard deployment loop → CI pipeline quality gate → Adversarial subagent review → Progressive production rollout.

Related quality assurance cases