Bring me the AI problem your team cannot solve with another prompt.
I work with organizations investigating difficult problems in AI architecture, agent systems, observability, provenance, uncertainty, and human-centered intelligence — the same problems documented publicly in Self-Mirrored and the research series.
No content strategy or engagement offer can guarantee a specific outcome. What follows is an honest description of the problems I investigate, how engagements are structured, and what you receive — not a promise of results.
Problems I am brought in to investigate
- —"We can see the final AI response, but we cannot reconstruct what happened around it."
- —"Our agent system has tools and memory, but no coherent lineage or audit model."
- —"Multiple agents disagree and our orchestration layer simply hides the conflict."
- —"Our AI always answers. We need reliable uncertainty and escalation states."
- —"We have a research hypothesis but need someone who can turn it into a prototype and evaluation."
Ways to work together
Fixed project fees or monthly retainers, scoped case by case. Ranges are directional starting points, not published pricing — see the fit call below.
Executive AI Strategy Session
1–3 daysA focused working session plus a concise decision brief — for teams that need clarity before committing to a larger engagement.
AI Auditability & Provenance Assessment
1–2 weeksArchitecture review, a trace/provenance gap map, and a prioritized risk/opportunity brief for one specific system.
Living Intelligence Research Sprint
2–4 weeksA defined hypothesis, an instrumented experiment, a working prototype, and findings — the flagship engagement, and the one most directly tied to public research.
Responsible AI Prototype Lab
4–8 weeksA working prototype, architecture, instrumentation, an evaluation plan, and a handoff — for teams that need a tangible product experiment, not another slide deck.
Research & Architecture Advisory
MonthlyRecurring review sessions, architecture critique, and experiment design — a standing technical partner rather than a one-off project.
Sponsored Research Program
6–16 weeksA defined research question, experiments, a technical report, a private briefing, and — by explicit agreement only — optional publication.
How the engagement works
- 1.Fit — A short call to confirm the problem belongs in this research domain — not a free architecture consult.
- 2.Define — A written research/architecture brief: scope, evidence requirements, deliverables, assumptions, and fee.
- 3.Instrument — Map or build the smallest system that can produce useful evidence about the question.
- 4.Investigate — Run the analysis, experiments, reviews, or prototype cycles that were scoped.
- 5.Audit — Inspect what actually happened, including negative or unresolved results — those are valid outcomes here.
- 6.Transfer — Deliver findings, architecture, prototype, roadmap, and a clear next-step recommendation.
If the problem is important, ambiguous, and technically difficult, that is usually the right place to start.
Start a Research Conversation