SELECTED WORK / SYSTEMS, AI & ASSURANCE

Ideas become useful
when they work.

I build integrations, operating views, and AI-assisted workflows that connect records to decisions. Start with a working public example, then explore the implementation and research behind it.

IMPLEMENTATION, RESEARCH & METHOD

Different questions.
Inspectable mechanisms.

01 / INDEPENDENT RESEARCH

Axiom

Memory that moves with you.

I keep a source journal outside a particular model, with derived retrieval indexes. I completed re-embedding, connected a fresh Claude instance, and tested retained memory and weighting in the replacement setup. The approach supports continuity through provider changes and project handoffs.

First-person private implementation account · Separate public context resolver and scripted model-swap illustration.

Explore Axiom and its public examples
Axiom conceptual model connecting context with observations, evidence, temporal state, evaluation, and provenance
Conceptual architecture illustration.
02 / APPLIED AI IMPLEMENTATION

AI Workbench

Bring requirements and source evidence into context.

I developed a Bedrock/Claude workbench with semantic retrieval and persistent context. Potential coverage gaps remain leads for human investigation.

Separate from the Navigator public demo and the Claude Code review described on the implementation page.

Read the workbench scope
Conceptual source systems connect to a context layer that supports explanation and decisions
Architecture model · The private workbench is available for discussion through a walkthrough request.
03 / ASSURANCE METHOD

Control Intent

Does the control achieve its purpose?

Translate an obligation into expected behavior, examine what evidence supports, and identify what to correct or validate next.

Authored method informed by professional practice. The access-review example is illustrative.

Follow the Control Intent example
Control Intent loop: risk, intent, behavior, evidence, gap, correction, and validation
Seven connected stages of assurance reasoning.
EXPLORE THE IMPLEMENTATION

Bring the system
you want to understand.

Pick a project and bring the question you’d ask if we were looking at it together. I’m happy to talk through the decisions, tradeoffs, and what the examples show.

Request a technical walkthrough

How this site represents evidence