Aztecscan

Aztecscan is a block explorer for the Aztec privacy-focused L2 chain. Over 21 months I contributed 678 commits to it across infrastructure and full stack — from the indexer pipelines that keep block data flowing, to front-end fixes, to the migration of the entire CI/CD pipeline off a hand-rolled Skaffold setup onto GitHub Actions + Kustomize.

What I worked on

• Indexing infrastructure running on Kubernetes, with Kafka at the centre of the data path and Postgres as the store.
• Progressive CI/CD migration — moved services one at a time onto the new pipeline, with explorer-ui-v2 as a live pilot while the old pipeline kept deploying the other seven.
• Migrating ingress from nginx to Envoy Gateway, one network at a time across the devnet / testnet / mainnet branches.
• Dropped-transaction false positives — a mempool-sync fix that added a suspected_dropped intermediate state with a 30s grace period instead of a crude timeout.

Agents, made safe for production

A thread running through the work: giving AI agents the context and guardrails to operate near production. That included a ~296-line devops.md agent persona encoding the repo’s Kafka conventions and release process, wiring @claude into GitHub code review, and delegating review fixes to a coding agent with self-verification.

Was it honest?

Two of the most attention-grabbing things an explorer could brag about aren’t mine to claim (protocol internals, etc.) — so this page sticks to the infrastructure, the migration, and the agent tooling, where the work is verifiable in the repository.