Deep dives into distributed systems, software architecture, and the engineering decisions behind production infrastructure.
Event-driven design decouples submission from processing, but it is not a default. A look at when EDA wins, when it loses, and what to design around.
A walk through token bucket, sliding window, and leaky bucket — and why the choice of algorithm is the smallest part of a distributed rate-limiting design.
Idempotency keys, offset commits, and poison-pill handling: the parts of a Kafka consumer design that nobody notices until they are wrong.
Hardening the front door of a backend: protecting third-party keys behind a proxy, JWT/OIDC verification, RBAC, and signed outbound IPNs with replay protection.
PostgreSQL indexing is one of the highest-leverage performance tools you have, but only if you read the query plan the planner actually generated.
Background jobs fail in interesting ways. Here is how to design a runner that fails safely, recovers automatically, and stops retrying when it should.
Bank-integrated payments demand more than idempotent endpoints — virtual accounts, dynamic fee calculation, IPN-driven state machines, and reconciliation that survives SFTP quirks.
Cross-cutting lessons from production backend work: ownership of shared libraries, changing one hot path at a time, observability before scaling, and contract-driven integration.