Infrastructure that stays up
Terraform-defined AWS and Azure infrastructure with Temporal orchestration, built for high availability on mission-critical financial services.
Context
Founding-engineer infrastructure has a specific failure mode: it gets built for the company that exists today and quietly becomes the reason the company that exists in eighteen months cannot ship. The work here was to build something a small team could operate now and a larger team could inherit later.
Constraints
- Mission-critical. Customers’ financial operations depend on these services being available, which sets the floor for redundancy and recovery.
- Two clouds. AWS for the core platform, Azure where customer or integration requirements demanded it — so identity, networking and deployment had to work in both.
- Small team. Every piece of infrastructure had to justify its operational cost.
What I built
Infrastructure as code, all of it. Terraform defines the AWS and Azure estate — networking, IAM, compute, data stores. Nothing important exists because someone once clicked a button in a console, which means environments can be reasoned about, reviewed and rebuilt.
Temporal for anything that takes longer than a request. Long-running, multi-step processes that must survive restarts, retries and partial failure are workflows rather than a chain of queues and cron jobs stitched together with hope. Failure handling becomes explicit and inspectable instead of emergent.
CI/CD on GitHub Actions. Tests, infrastructure plans and deploys run on every change, so the path to production is the same path every time.
Decisions
Temporal over a queue-and-cron architecture. Queues make the happy path easy and the failure path invisible. For workflows where a partial failure means a financial document is half-processed, durable execution with explicit compensation was worth the operational overhead of running Temporal.
Terraform from day one, before it was painful. Adopting IaC early is cheap; retrofitting it onto a live estate is not. This was a deliberate up-front cost.
Outcome
High availability for mission-critical services, and an estate that can be reviewed, rebuilt and handed to new engineers — which became the foundation the SOC 2 work was built on.