Systems

I build execution-governed systems for exploratory computation and AI workflows. The focus is reliability under iteration: determinism, auditability, and fail-closed control.

AUFG Research System

AUFG (Alexandrian Unified Field Geometry) is run like a lab-grade computational program: every experiment is state-locked, reproducible, and falsification-first. Runs produce structured artifacts (meta, logs, failures, summaries) under deterministic authority rules, so results can be replayed exactly and compared across revisions. The goal is not “pretty outputs,” but stable measurement: when something changes, the system proves why.

Fail-Closed Execution Governance

A run is considered invalid unless correctness is explicitly established. Preflight validation and runtime enforcement convert silent failure modes into explicit invalid runs, preventing “successful but incorrect” outputs from contaminating results.

Artifact-First Authority

Execution produces authoritative artifacts for every attempt—successful or not. Metadata, failure logs, and run directories are created or reserved before computation begins, enabling traceability and post-hoc auditing at scale.

Deterministic Seed and Interface Authority

Determinism is treated as an execution invariant. Seeds are derived from canonical inputs, not runtime randomness. Parameter transmission and callable interfaces are enforced at runtime to prevent silent drift.

Corpus-Governed Novelty Checking

New ideas are evaluated against a canonical indexed corpus of prior patents and papers. The workflow is designed to reduce overlap risk and accelerate iteration: index → overlap scan → verdict → design-around.

Exploratory Research Infrastructure

Long-horizon experimentation requires governance beyond unit tests and CI. I design systems that preserve development velocity while maintaining reproducibility and interpretability across runs.