Serverless systems on AWS and Azure, designed and built by the same person. Terraform underneath both, so the infrastructure is versioned, reviewable, and portable instead of clicked together in a console.
An architecture document is only worth what gets deployed from it. Engagements here cover both halves: the design that gets agreed on, and the working system that comes out the other end.
Assessment of what you are running now and a practical plan for what it should be, sized to your institution rather than lifted from an enterprise template. Event flows, service boundaries, failure modes, and cost shape decided before anyone writes code.
Functions, queues, event routing, and scheduled workloads written, deployed, and handed over running. Terraform for every environment, logging and alerting wired in from the start, and documentation that lets your team take it from there.
Capacity is defined from day one as scope categories plus a 2 hour response SLA. Generous by design. Work outside these categories is scoped as project work.
A federal satellite data program ingests files continuously from ground stations and international partner servers. The pipeline runs on Lambda, SQS, and EventBridge, pulling files that reach 400 MB and routing them by product type into storage for downstream processing.
The work covered both the architecture and the code. Large files were timing out because the chunked download path silently never engaged in production, so transfers ran single stream at a fraction of the available bandwidth. A local disk chunked download path fixed it. Range request support had to be forced for an upstream server that honored range requests without advertising them. Download concurrency was benchmarked rather than guessed: throughput climbed steadily through 80 parallel connections and fell off sharply above that, which set the safe ceiling.
Separately, an audit of the product routing configuration raised the match rate from 21 percent to 41 percent across a dataset of more than a million records, by correcting timestamp handling, instrument grouping, and pattern ordering across hundreds of product types.
Client and program not named at their preference. Further detail available under NDA on a call.
Asset size sets the public price band. Discovery places you inside it based on environment complexity, integration surface, and change velocity. Not bundled into a DL4 engagement.
Discovery call $850 flat fee, separate from the retainer. New retainers start with a three month initial term, then true up. Project work outside retainer categories $160/hr.
An $850 flat fee review of your current architecture and what it would take to run it well. Scope categories, band placement, and next steps confirmed before any retainer begins.