RESOURCE MAP

Choose by the decision you need to make.

Every resource states its starting level, output, tools, access, version, and evidence boundary before you open it.

01Start

Run one transparent baseline.

02Build

Preserve the engineering contract.

03Evaluate

Test leakage, error, and failure.

04Govern

Define evidence and authority.

Public resources open without email. The subscriber Blueprint arrives by email after confirmation. The beginner lesson and downloadable teaching examples are free to use without subscribing.

01 / Start

Build the first transparent result.

Recommended first step · Runnable teaching auditv0.2.1-previewPublic

Power-Electronics ML Baseline Starter Kit

Two synthetic teaching exercises for auditing grouped validation, direct-label leakage, and a model-versus-engineering-rule comparison.

Outcome: A reproducible rule-versus-ML decision, followed by a grouped proxy baseline that remains only a validation candidate
Tools: Windows runner or Python 3.13.5; NumPy 2.2.3; pandas 2.3.3; scikit-learn 1.8.0
Starting level: Power-electronics experience; little or no Python assumed
Format: Two notebooks + two CSVs + runner + verified outputs + checklist + licence
Evidence boundary: Synthetic data only. Exercise 1 contains direct-label leakage and prefers the exact rule; exercise 2 uses proxies and still does not estimate hardware performance.
Updated: 2026-08-26

Open the starter kit

02 / Build

Turn engineering work into traceable artifacts.

Published evidencev1.0.0 evidence releasePublic

Field Note 001: define what pass means

A rebuildable LLC workflow that preserves candidate records, gate results, replay evidence, and a separate method failure.

Outcome: A traceable evidence loop and explicit acceptance boundary
Tools: Browser; optional Python or PowerShell quickstart
Starting level: Power-electronics workflow and simulation experience
Format: Long-form case + public artifacts
Evidence boundary: One operating point and a simplified equivalent model; no hardware approval or solver equivalence.
Updated: 2026-08-22

Read Field Note 001

Published artifactv1.0.0Public

LLC workflow Quickstart

Run a bounded 20-candidate workflow and inspect the output integrity, decision fields, plots, and replay contract.

Outcome: A replayable workflow run, not a trained ML model
Tools: Windows launcher, Python, PowerShell, Jupyter, or Colab
Starting level: Converter design; no code editing on the Windows route, but a working Python launcher is required
Format: ZIP with scripts, notebook, checks, and outputs
Evidence boundary: The exported table is not a qualified training dataset; the package demonstrates an evidence workflow.
Updated: 2026-08-22

Download Quickstart v1.0.0

03 / Evaluate

Test the claim where it can fail.

Working noteWorking versionPublic

Six checks before trusting a surrogate

A plain-language acceptance path covering baselines, unseen designs, residuals, unfamiliar inputs, uncertainty, and abstention.

Outcome: A review contract for deciding when a surrogate may be used
Tools: Model results and engineering review
Starting level: Engineering-model evaluation
Format: Working note + checklist
Evidence boundary: Thresholds remain decision-specific and require independent evidence; no universal confidence threshold is claimed.
Updated: 2026-08-12

Read the acceptance checks

Published artifactv1.0.0Public

LLC evidence archive

The deeper 200-candidate archive for inspecting records, manifests, verification, replay, and the preserved method failure.

Outcome: A fuller audit trail for the Field Note 001 campaign
Tools: Python and archive inspection
Starting level: Comfort reviewing structured engineering evidence
Format: Versioned ZIP archive
Evidence boundary: The archive does not add multi-point, higher-fidelity, or hardware validation.
Updated: 2026-08-22

Download the evidence archive

04 / Govern

Define evidence and human authority.

Subscriber resourcev1.2Subscriber email

Power-Electronics Workflow Automation Blueprint

A seven-module governance structure for decomposing expert work, defining evidence, and preserving human review gates.

Outcome: A reusable workflow, evidence, and acceptance canvas
Tools: PDF reader
Starting level: Engineers designing repeatable technical workflows
Format: PDF delivered through the confirmed-subscriber welcome email
Evidence boundary: The download link is sent after email confirmation. This resource supports engineering review; it does not approve a design.
Updated: 2026-08-26

Read about the subscriber Blueprint

Not sure where to begin? Use the public synthetic baseline first; subscription is optional.

Open the recommended first step