Herman Boma

My primary obsession is AI agents and agentic systems — how they are built, deployed, and composed. Not as a theoretical exercise, but as a design problem shaped by memory, context engineering, tool design, and human-in-the-loop constraints.

Before the current wave, I built a career around data engineering and cloud infrastructure — AWS, HashiCorp, Cisco networking, data mesh, dbt, and streaming architectures. That background keeps me interested in systems that must survive contact with real infrastructure and real organizations.

I am drawn to thinkers who name things: people who coin frameworks, expose antipatterns, and give fast-changing domains vocabulary sturdy enough to think with.

I save aggressively and engage selectively. The library is an aspiration buffer. But when a source earns serious attention, it gets read with intensity. Dancing in the Glory of Monsters has 460 highlights; the median source has 9.

My reading collapsed between 2017 and 2019, then exploded. The catalyst was agentic AI: suddenly, everything I had read about systems thinking, architecture, governance, and failure analysis became relevant in a new context.

I gravitate toward practitioners over commentators, conceptual density over length, and the intersection of architecture, philosophy, and technical writing.

I help organizations navigate the intersection of AI agent architecture, data engineering, and enterprise systems design.

• Designing agentic AI systems: tool architectures, context strategy, human-in-the-loop workflows.
• Data platform strategy: mesh architectures, pipelines, streaming, and operational design.
• Technical strategy: bridging what AI can do with what organizations can safely absorb.

Reading: trust thresholds, perception gaps, context engineering, and the messy frontier between what agents can do and what users think they can do.

Building: this site as both editorial publication and archive interface.

Thinking about: why metrics design is the binding constraint on autonomous research, what it means to make the world safe for imperfect humans, and how every research organization eventually turns into a `program.md` problem.