Designing for failure before it happens →
Good architecture reduces the blast radius and improves recovery confidence long before an incident hits.
Articles
Articles explain the architecture and trade-offs behind CloudJi projects, from the evolving Core platform to Fundamental infrastructure and Individual projects.
The evolving AI infrastructure platform
Good architecture reduces the blast radius and improves recovery confidence long before an incident hits.
North–south entry paths, private workloads, NAT trade-offs, VPC endpoints, security groups, and DNS.
S3 events, queues, workers, retries, DLQs, idempotency, duplicate delivery, and backpressure.
Indexes, mappings, inverted indexes, analyzers, BM25 scoring, shards, and replicas—explained from an infrastructure viewpoint.
The request flow, temporary permission, content validation, expiration, and safe upload boundaries.
Why duplicate events are normal and how workers avoid corrupting state while safely retrying.
Short-lived environments, budgets, teardown plans, cost-heavy services, and what to simulate locally.
Standalone projects and experiments
A practical framework for structuring a lab around one objective, clear prerequisites, verification, troubleshooting, and cleanup.
Foundational knowledge and infrastructure concepts that support multiple projects
The value of observability is not collecting more data; it is creating clearer operational decisions.
Good infrastructure code grows with the team when structure and dependencies are deliberate.