I build systems that understand before they act.
Distinguished engineer. Ontology, context engineering, agentic development, and knowledge systems, designed across security, infrastructure, data, scale, networking, and architecture. The deliberate pursuit of simplicity in complex domains.
What I work on
Two paths, one idea
Better action depends on seeing the system more clearly. One path makes the real state of organizational work visible. The other makes meaning, evidence, and constraints visible to software and AI systems.
Working Faster
You change a system by changing yourself. Your expectations, what you measure, and how you articulate the cost of a delay. Shorter timelines, not longer hours.
- Start here
- "On Track" Is Not a State
- Core concepts
- Calendar Time and Touch Time. Wait Time. Opaque Status. Progress Shapes. The Second Question.
- Current focus
- Publishing the framework one piece at a time, and building a field guide from the vocabulary.
Context Engineering
Systems that organize meaning, evidence, and constraints before people or AI act.
- The idea
- Engineer the information environment rather than adding rules. Ontology, retrieval, evidence, and structure carry more weight than instructions.
- Core concepts
- Ontology as architecture. Context before action. Knowledge structure. Structured prompt-driven development. Agent and system architecture.
- Current focus
- Designing and testing my own approaches to retrieval accuracy, cost, and risk, rather than only applying published patterns. The applied work is largely proprietary. The models behind it come from self-directed experiments, and those are what I can share.
Point of view
How I think
Ontology as architecture
Systems inherit the shape of the concepts they model. Get the ontology right and the code writes itself. Get it wrong and no amount of engineering saves you.
Context over instruction
The best systems don't need more rules. They need better context. Engineer the information environment and let intelligence emerge from structure.
Simplicity is the residue of understanding
Complexity is not a problem to manage. It is a symptom of incomplete comprehension. Simplify by understanding more deeply, not by hiding more aggressively.
Speed comes from anticipation
The only way to move fast is to know what every team will ask before they ask it. Security, data, infrastructure, architecture, the C-suite. Their questions are design constraints, not gates to clear later. Anticipate them and you stop trading one pillar against another.
Featured thinking
Writing
Working Faster is about where delivery time actually goes. Published essays, newest first.
"On Track" Is Not a State
"On track" is a claim about the date. It is not a description of the work. Twenty minutes of real effort and a week in someone else's queue sound exactly the same, and the calendar moves either way.
You will learn how to hear the difference between a status and a state, and the one extra question that gives a wait an owner and a next signal.
Complaints Don't Travel Upward. Costs Do.
A complaint names a feeling, and a feeling can only be sympathized with or dismissed. A cost names a wait, a size, and an ask, in the terms the person deciding already tracks.
You will learn how to turn a recurring friction you do not own into a number someone can act on, and why keeping the friction can still be the right answer.
Proof of work
The arc
The ideas above did not arrive as opinions. They came out of building. Most of this work is applied inside organizations and is not public, so what follows is the sequence rather than a set of links.
Knowledge Catalog
How do you make what an organization knows queryable, composable, and alive? Started here.
Thinking Tool
If knowledge is the material, thinking is the operation. Built tools for structured reasoning that make the invisible work of understanding visible.
AI Workbench
Context-rich environments where AI and human reasoning compose. Not chat. Workbench. The right information at the right time.
Working Faster
Turned the lens on the craft itself. What does high-leverage engineering look like when you stop confusing motion for progress? This one became public writing.
Read the essaysStructured Prompt-Driven Development
Decomposing work into composable concept-prompts. An ETL for knowledge. Raw captures ideas as they arrive. Lenses transform them into structured artifacts. Deliverables derive from lenses on demand. Regenerate anything without losing information.
Proving Ground
An end-to-end experiment applying the full stack. Ontology, context engineering, composable prompts, and raw-lens-deliverable knowledge capture. Validating that the thinking holds under real delivery pressure.
About
Jeff Olsen
I work across security, infrastructure, data, scale, networking, and architecture, and translate between engineering teams and the C-suite. People, process, and technology is an overused phrase and still the right one. Most hard problems live in the seams between the three.
This is my hobby, my passion, and my vocation. The side projects, the writing, and the paid work are the same activity, which is why the models here come from practice rather than theory.
Subscribe and connect