Jonathon W. Marshall
Building platforms for the age of agents
I'm a platform engineer with more than twenty years building and running infrastructure, most recently focused on taking AI systems from prototype to production. The path ran from DevOps and multi-cloud platforms to the security work and guardrails that let teams adopt agentic tools without losing control of their data.
What I care about is systems that last: infrastructure you own rather than rent, data that stays sovereign, and software built on durable foundations instead of the churn of corporate web trends. That conviction shapes the tools I reach for and the way I think about portability and failure.
Away from the keyboard, I'm usually making something physical or out in the backcountry with a canoe and a film camera.
Contact
hello@jwmarshall.devCurrently Learning
I believe Elixir, and the BEAM virtual machine it runs on, is one of the most capable foundations available today. The BEAM provides lightweight processes, message passing, supervision trees, and built-in pub/sub natively, the things most stacks bolt on as separate task queues, message brokers, and in-memory stores, each adding operational overhead and failure surface. Moving those into the application itself means deeper integration with far less to manage. Phoenix and LiveView go further, delivering real-time interfaces over WebSockets without a separate JavaScript framework or API layer, collapsing the frontend-backend divide into one language and runtime. Ash adds a declarative, resource-based layer that generates authorization, filtering, pagination, and API endpoints from concise definitions instead of boilerplate. Elixir also ranked as the second most admired language in the 2024 Stack Overflow Developer Survey, and its consistency and strong documentation make it well-suited to AI-assisted development.
Interests