How I think, design, and build.
I turn abstract ideas about trustworthy AI into small, reviewable artifacts and applied research. Public repositories show the method; the research below shows the range. Most production work is private — these are the clean-room versions that show how I work without exposing private systems.
How I work
The same loop runs on every project — the difference between a claim and something you can trust is whether each step actually happened.
Follow the pattern
Start from a real problem and pull structure across disciplines — ethics, cryptography, cognitive science, systems design — instead of staying inside one field.
Make it bounded
Turn the idea into an architecture with explicit boundaries: what it may do, what it must refuse, who stays in the loop, and how a decision is recorded.
Ship the small version
Build the smallest real thing that runs under real constraints — a scorer, a schema, a runner — rather than a slide about what it could do.
Test & record
Attack it, gate it, and leave a reconstructable record. If it can't survive its own review, it doesn't ship.
Publish honestly
State the claim, show the evidence, name the limitation. Restraint is the credibility.
Governance & evaluation
The public core: making governance claims inspectable instead of aspirational.
AI Governance Benchmarks
A two-layer evaluation stack for agent boundary safety and human sovereignty: synthetic inputs → a deterministic rule-based scorer → reproducible reports → passing tests, with CI and an anti-overclaim language guard.
View repository → AuditabilityAgent Action Audit Template
Schema-backed synthetic agent-action receipts with safe vs. blocked-action examples and human-review checkpoint metadata, validated through JSON schemas and pytest.
View repository → Human oversightHuman–AI Governance Lab
A reproducible workflow for lightweight risk classification, human approval gates, bounded delegation, and auditable, reconstructable decisions.
View repository →Applied research
Longer-form work on how governed, sovereign AI systems actually get built. Selected papers — the strongest are on Writing.
Sovereign AI
Distributed cognition on consumer hardware with a governed boundary that keeps local models out of the reasoning path — cloud-grade capability with zero data egress. Honest about its own limits.
Read the paper →Verifiable Self-Governance
A formal approach to safety in adaptive autonomous systems: specifying and checking governance properties so a system's constraints can be verified, not just asserted.
Explainability, Auditability & Data Sovereignty
A framework for advanced AI systems that keeps decisions explainable, records reconstructable, and data under the owner's control.
Explorations
Where I follow a pattern into new territory — bounded, honestly caveated, and useful for showing how I think.
Active Inference Primer
A minimal educational primer on prediction, uncertainty, and agency for anticipatory cognition in autonomous agents.
View repository → AlignmentEudaimonic Alignment
Research notes exploring AI wellbeing as an alignment strategy, drawing on Aristotelian philosophy and mapping knowledge traditions to modern governance.
View repository →UNA Distributed Cognition
How reasoning, memory, and perception split across nodes while a governed boundary keeps the reasoning path clean — the architecture behind UNA.
GeoAIF — Edge Seismic Intelligence
A design for real-time seismic detection and early warning over mesh sensor networks and edge AI, without centralized infrastructure.
THRUM — Swarm Coordination
Exploratory work on coordination and learning under constraints in multi-agent, swarm-like environments.
Quantum + AI Experiments
Simulator-first experiments at the quantum/AI boundary, with explicit claim limits — clearly labeled non-production.
View repository →The discipline behind all of this comes from UNA — the deterministic, governed AI system I build and run. Its governed change pipeline, capability boundaries, and auditable receipts are the private counterpart to the public work above. See how UNA's governance works →
Want the full research corpus or a specific paper? Get in touch — several papers are in preparation and available on request.