Selected technical work
The selection begins with operational AI, private local systems, evidence-controlled research and original mathematical computation. It then broadens into peer-reviewed methods, industrial integration, scientific software and experimental products.
Four proof anchors
Correspondence Matrices makes an original representation runnable and checkable. Nexus, Local Multi-Agent AI and Paper Workbench show operational control, private analysis and research provenance in implemented systems.
Watch CM Benchmarking Video →Algorithms & structural reasoning
Original method and measured computation.
An original Boolean representation taken through public code, independent checks and a benchmark audit.
Public implementation · Published audit · Authored paper
Watch project video →AI operations & exact action authorization
Useful agents. Bounded authority.
A human-governed agent platform for real email and inbox workflows, bounded research and controlled external integrations.
Working platform · Provider evidence · Auditable controls
Watch project video →Applied AI / engineering automation
Evidence-grounded engineering analysis.
A local, air-gapped framework combining technical-document retrieval, hybrid search, deterministic NLP and configurable multi-agent workflows.
Air-gapped system · Reproducible retrieval evidence · Public role context
Watch project video →Evidence-controlled research & publishing
Research claims with an audit trail.
A public research workbench that keeps sources, claims, human decisions, audits and exports connected by provenance.
Public implementation · 101 offline tests · Evaluation reports
Further engineering & research evidence
Peer-reviewed machine learning leads into industrial systems, numerical thermophysics, computational neuroscience and two public product experiments.
Watch project video →Explainable ML & measurement
SHAP-guided item retention.
A first-authored, peer-reviewed study investigating whether model-derived item importance can support shorter psychometric measures.
Peer-reviewed study · Public data and analysis
Systems integration / industrial data acquisition
PC Acquisition Host ↔ Siemens S7-1500
NASA Ames PROFINET and Siemens S7-1500 data-acquisition systems integration involving HMS/Ixxat INpact, GSD/GSDML, embedded C and PLC process-data mapping.
Bounded role account · Contract map · Public context
Scientific computing / applied mathematics
Arc-jet calorimetry translated into software.
Scientific-computing implementation of a quasi-1D conical-frustum inverse heat-conduction model for NASA Ames arc-jet calorimetry using LabVIEW.
Published NASA method · Software translation · Public role context
Computational neuroscience & research methods
Anxiety dimensions. Network differences.
A co-authored 2025 resting-state fMRI study using independent components, dual regression and FSLNets to study anxious apprehension and arousal.
Peer-reviewed study · Published method · Role context
Mathematical visualization & product systems
Equations become a visual medium.
A public mathematical-art platform connecting equations, rendering parity, composition scoring, preference learning and recoverable encoded messages.
Live product · Public gallery · Cross-renderer work
Simulation, rules & production tooling
Build a system. Then disrupt it.
A live strategy-card-game project supported by simulation, constrained visual-language research, a digital implementation and reviewable media workflows.
Live product · Simulator evidence · Published mechanics
Independent research programme
A mathematical programme on self-reference.
A bounded account of B-Theory, Brian’s long-running independent programme connecting logical states, topology, homology and continuous change. It is presented separately from the numbered project accounts; Correspondence Matrices is the programme's implemented computational offshoot.
Authored programme · Working papers · Public research map
Explore the B-Theory programme →How to read the work
What was difficult, uncertain or worth testing?
What representation, model or system made the question concrete?
What can a technical reader inspect, rerun or read?
What does the work not establish?
Work is not the timeline
These accounts are selected evidence clusters: substantial enough to explain the difficult question, implementation and limits. The CV remains the complete chronological record, including earlier institutional research that does not yet have a public project account.
See the complete project and experience record →
