Work

What It Is

A complete backend system for managing recurring subscriptions—handling sign-ups, automated billing, renewal scheduling, and notifications. Behind the scenes, it utilizes a robust asynchronous worker pipeline (Asynq/Redis) to process thousands of lifecycles without human intervention, all monitored by an enterprise-grade telemetry stack.

Why It Exists

Subscription state bugs cost real money. When billing systems are built as an afterthought, users get charged after cancelling or stuck in ghost states. I built this architecture to mathematically prove that invalid states can be made impossible at compile-time, and to demonstrate that defensive testing and distributed tracing should be first-class architectural citizens.

Key Decisions

• Interface Segregation — Service interfaces are strictly split into External (API) and Internal (Workers), physically preventing HTTP controllers from triggering background logic.
• Database-Agnostic Transactions — Transactions are executed via an injected TxnFn closure, allowing hermetic service unit tests without importing MongoDB drivers into the domain.
• Spoofing-Resistant IP Extraction — The system traverses X-Forwarded-For strictly right-to-left to extract the true client IP, actively ignoring attacker-controlled prepended headers.
• Asynchronous Task Chaining — Domain mutations and side effects are decoupled. If the SMTP server crashes, Asynq safely retries the email task without re-running the billing transaction.
• Deterministic Clock Injection — Time is injected via a NowFn closure rather than relying on global wall-clock drift, enabling mathematically perfect boundary testing for subscription expirations.

Trade-offs & Scope Management

Senior engineering requires deliberate architectural constraints. To maintain boundary integrity, I explicitly enforced:

• Modular Monolith over Microservices — Strict domain separation ensures it can be split cleanly later, but a single deployment binary avoids premature network latency and orchestration overhead today.
• Hermetic Reliability over Speed — Implemented strict trace-correlated logging and isolated Testcontainer database setups. It takes slightly longer to build, but completely eliminates flaky tests.
• Banned Cross-Domain Reads — Services are physically restricted from querying other domains' repositories to prevent schema coupling and enforce strict data boundaries.

What It Is

An end-to-end URL shortening service. Instead of relying on heavy web frameworks, I built this to test the absolute limits of Go's standard library (net/http), explicit middleware orchestration, and containerized deployment pipelines.

Why It Exists

I wanted to build something real, not just follow tutorials. The goal was to own the full cycle: design → build → deploy → maintain. By keeping scope small, I could focus on doing each part properly rather than building half of something impressive.

Key Decisions

• Zero-Magic Routing — No heavy web frameworks. Handlers and middleware are explicit, predictable, and rely solely on Go's standard library and chi.
• Interface-Driven Storage — The UrlStore interface completely abstracts MongoDB. The storage layer can be swapped without touching a single line of business logic.
• Release-Triggered CI/CD — Engineered a fully automated GitHub Actions pipeline. Code is never deployed on merge; it is deployed on tagged releases, providing a clear audit trail and immediate rollback targets.
• Optimized Containerization — Utilized multi-stage Docker builds to dramatically reduce the final image size and attack surface.
• Graceful Shutdown — Server listens for OS signals, finishes in-flight requests, then closes connections cleanly.

Trade-offs & Scope Management

Senior engineering requires knowing what NOT to build. For this architectural study, I explicitly scoped out:

• Rate Limiting & Caching — Premature optimization for a deployment-focused study.
• Authentication — Unnecessary for the bounded context of this specific API.
• Analytics — Scope creep. The goal was lifecycle ownership, not feature bloat.