Patrick Elmore
Engineering manager building the AI-driven systems that build software, at enterprise scale. The substrate of software engineering is shifting, and most of the high-leverage work is moving up the stack, from writing implementations to designing the systems that produce them. I write about the practice this shift actually requires.
Background
I lead a software engineering team operating at the forefront of AI-assisted development, with a year of production experience using AI coding tools as the primary substrate for how we plan, implement, and validate work. The methodology is spec-driven: specification quality is the primary variable that determines what the AI produces, and the discipline of building good specs is most of what makes the system work.
The team's process is codified in a planning repository that I designed and continuously refine: agents, skills, hooks, and validation patterns calibrated to the codebase and the team. The approach is evidence-based, grounded in a dataset of classified incidents, validations, and named constructs that inform how the system evolves. The methodology has spread to additional teams through voluntary adoption.
Outside the AI work, I architect and maintain distributed, fault-tolerant systems on Azure, including service-bus-driven architectures designed for resilience under partial failure. I build the supporting application layer (Azure Functions, ASP.NET APIs, SQL Server) with attention to query design and performance at scale, and own end-to-end CI/CD in Azure DevOps.
What I find most interesting about this space is that durable capability is built, not adopted. It emerges from teams doing the deliberate work of developing their own AI-driven processes against their own codebases, constraints, and failure modes.
Based in Indianapolis.