Projects

Selected backend systems focused on event-driven design, reliable data processing, and practical scaling trade-offs.

Order Tracking — Event-Driven Ingestion with Kafka

A Kafka-backed order ingestion path that absorbs bursts without coupling HTTP latency to database processing, while keeping state-transition rules outside messaging and persistence concerns.

Decision

Order ID is the Kafka key, so events for one order share a partition and preserve per-order ordering while unrelated orders continue in parallel.

Limitation

The flow is not end-to-end exactly once: there is no durable event identity for database deduplication or a dedicated safe replay path yet.

Project overview Engineering write-up Repository

Snapshot Ingestion — Resilient Coordination with PostgreSQL

A snapshot-ingestion service that isolates unreliable provider fetches from the search path, stages streamed XML durably, and recovers worker progress through visible PostgreSQL state.

Decision

Because the source is a periodic snapshot and PostgreSQL is already canonical storage, the database also owns durable staging instead of adding broker operations before the workload requires them.

Boundary

Advisory locks serialize provider fetches and SKIP LOCKED distributes staged work; the shared database remains the scaling limit if contention, fan-out or replay requirements grow.

Project overview Engineering write-up Repository