/// How I Work

A competency list proves nothing. Here is the same claim with receipts attached.

Every capability below carries one instance from the work — a part number, a measured delta, a decision and what it cost — and links to the page that documents it. Where a number cannot be pinned to an artifact, it is not on this page. Arranged by the physics each practice governs, the same four domains used to arrange the work itself.

The method field

Matter & HeatMotion & FaultData & AILight, Sound & Perception
20 nodes · hover to identify · click to open

Matter & Heat

High-Stakes Physical Architecture

Motion & Fault

Kinetic Diagnostics & Combustion Logic

Data & AI

Systemic Sovereignty & Code Logic

Light, Sound & Perception

Sensory Architecture & Waveform Physics

20 practices · hover to identify · click to open the evidence

Matter & Heat

High-Stakes Physical Architecture

The governance of matter behaving at boundary conditions — thermal budgets, tolerance stacks, and structures that fail visibly when the physics is not respected.

Motion & Fault

Kinetic Diagnostics & Combustion Logic

The machine as a legible system: input, process, output, failure mode, root cause. Learned on outboards and tugboats, matured on engines under load.

Data & AI

Systemic Sovereignty & Code Logic

The architecture of persistent state and identity — durable, auditable, operator-owned, and designed to survive the container it runs in.

Light, Sound & Perception

Sensory Architecture & Waveform Physics

Engineering at the boundary between the physical world and human perception — audio waveforms, projected light, captured time.

Design

Fence the failure modes before anyone cuts steel.

  • Thermal & EMI architecture

    Validated active-cooling fallbacks on the C|24, then removed the heat source instead — re-architected 4U to 3U and moved the PSU to an external brick, shipping a silent fanless console that held its 10–35 °C window.

    Evidence: c24

  • Tolerance, stack-up & mechanical integration

    Authored the Data Control Drawing protocol on the C|24: 50+ binding geometric contracts overlaid against the 3D master and rejected on 0.5 mm variance. 100% mechanical fit on the first physical build of 19 PCB assemblies.

    Evidence: c24

  • Mechanism & actuated-system design

    Lead mechanical architect on a cobotic food-assembly makeline coordinating 70+ actuators at ±2% portion precision and 350 meals per hour.

    Evidence: makeline

  • Human factors & class-A surfacing

    Characterised the Glyph headband against a documented trilemma of retention, acoustic seal and pain — settling at 7.5 N clamp force on 0.8 mm hard-rolled stainless, then compensating the acoustic leak electronically rather than pretending the compromise away.

    Evidence: avegant-glyph

  • Haptic & kinematic tuning

    Preserved the premium 'scrub' feel of a $10,000 console on commodity parts — a Bourns EM14 jog wheel with six exposed 0.42 mm leads got a pre-terminated harness spec and a custom surround (ECO 13082) rather than a cost-up.

    Evidence: c24

  • Human-engineering standards & anthropometric fit

    Sized the Glyph headband, IPD range and nosepiece against military human-engineering criteria (MIL-STD-1472F, MIL-STD-1787C) and aircrew anthropometry — Bitragion-Coronal Arc, Nasal Root to Wall, Menton Projection. The forensic audit also records what that cost: 1960s US aviator surveys assume a long, narrow Western head profile, so short-and-wide profiles fit poorly. Uneven weight distribution and nosepiece discomfort across a large share of the global market, traceable to the data source rather than to the geometry.

    Evidence: avegant-glyph

Diagnosis

Find the constraint. Not the symptom, and not the person.

  • Failure-mode analysis & root cause

    Traced a 100% field-failure mode in a disc-changer mechanism through 3,000,000+ logged cycle events to a friction-coefficient drift.

    Evidence: m700

  • Process forensics

    Traced 2.50 mm of ABS 'banana' warp to heat-facilitated creep in a paint-cure cycle — parts baked flat on wire racks, below glass transition, sagging under their own weight. Not a material problem. A fixturing problem.

    Evidence: c24

  • DOE & test-method design

    Rejected the vendor's proposed fix and ran a comparative cure study instead — Method A vs B vs C — then codified the winner as a permanent manufacturing spec rather than a one-off rescue.

    Evidence: c24

  • Accelerated life & destruction testing

    Glyph's eyepiece arms carried a 3,000-cycle requirement and got worse as tooling matured: 10% seizure at T1 (500 cycles), then 40% — two of five units — failing at just 250 cycles in T6, on the eve of ramp. Root cause was internal cable wear, found because the test ran every tool revision.

    Evidence: avegant-glyph

  • FEA-correlated validation

    Optimized the Glyph headband through a Central Composite Design FEA study and correlated it against ANSYS stress-life binders and physical spring-rate characterization — simulation used to bound a real part, not to decorate a review.

    Evidence: avegant-glyph

Production

The design is not finished until the line can build it at yield.

Governance

If it is not in the record, it did not happen and it will happen again.

The part that is not on the résumé

The record is the deliverable. A fix that lives in someone's memory is a fix that gets re-discovered by the next team at full price — so the vertical-hang cure fixture became an ECO, the geometric contracts became a released drawing set, and the honest compromise on clamp force got written down as a compromise. That habit is why there is thirty years of this to show at all.

The same discipline now runs the infrastructure behind this site: raw program archives compiled into one sourced record, projected to these pages as a read-only render target so a page cannot drift from its evidence. That machinery is documented in the colophon. It is the instrument, not the product.

Tell me what's breaking