Shawn McBurnie

External validation: AICred: Rank 1 on the 2026 leaderboard, 10.0/10 (assessed 2026-07-21). 2025 board: #2 of 248, 10.3/10, rated Elite under that year’s rubric.

Co-founder of the Veridi (fact verification) / Pragma (policy synthesis) / Praxis (action synthesis) methodology suite: three aligned frameworks for what is true, what works toward a chosen outcome, and what you can do from where you stand. The Veridi web app is in invite-only beta as of April 2026; Pragma and Praxis run inside Veridi and also ship as standalone Claude Code skills. The AI Reliability Layer (evaluation, verification, and agent reliability under adversarial conditions) and UI are my work, architected and verified.

20+ years in software engineering, with the last 15+ concentrated in nonprofit engineering, enterprise design systems, and applied AI. Twin Cities.

Currently salaried full-time as Lead Engineer at BI Worldwide; open to senior-to-staff IC roles in Applied AI Engineering and adjacent tracks. See availability for fit, locations, and contact.

Shawn McBurnie is an Applied AI Engineer based in Minneapolis with 20 years of independent system engineering experience, now operating at the leading edge of agentic AI development. He builds production-grade AI systems with rigorous evaluation discipline, including a 7-agent FastAPI pipeline backed by Brier-calibrated specifications and a 1,500-line product spec currently in invite-only beta. Shawn authors Claude Code skills with formal test contracts, architects multi-agent workflows with governance frameworks built in from the start, and applies the same evaluation-methodology rigor to commercial software development gated by 10,000+ automated tests. His approach treats AI externalization as an engineering problem—designing artifacts precise enough that capable developers can ship from them without hand-holding. Shawn brings expert-level prompt architecture, agentic system design, and a rare combination of builder instincts and evaluation discipline to any team serious about moving AI from prototype to production.

AICred assessment: Rank 1, 10.0/10 on the 2026 board (assessed 2026-07-21); 2025 board: #2 of 248, 10.3/10. Quoted as generated; the product spec it cites at 1,500 lines stands at 1,968 as of v1.4 (see work).

Where the body of work shows this:

• Veridi: 7-agent FastAPI + Anthropic SDK pipeline, Brier-calibrated, 13 documented gaming countermeasures, public validation report. Invite-only beta as of April 2026.
• NetterTech Events: 5-plugin commercial WordPress suite gated by 10,000+ automated tests, PHPCS + PHPStan + PHPUnit pre-commit gates, OWASP 2025 posture, WCAG 2.2 AA throughout. Free core live on WordPress.org.
• Over 80 Claude Code skills authored with declared test-contract patterns, covering multi-agent orchestration, evaluation governance, and verification methodology. Two are published as open source (MIT) at claude-a11y-skills, test-contracts intact.

Recent and current organizations: BI Worldwide · Anaplan · National Marrow Donor Program (now Be The Match) · Nerdery · Clockwork · Celtic Junction Arts Center · Ireland Network Minnesota.

If you are running production AI in 2026, you are carrying runtime exposure that does not show up in benchmark accuracy: verification, calibration, adversarial robustness, and recovery. My Reliability Layer manifesto traces how I frame that gap. The work below is what I have built against it.

Whether the person in front of the system can actually perceive its output is one of those exposures, not a separate program. Twenty years of accessibility engineering taught me the same lessons production AI is learning now, and Accessibility is a reliability concern makes that case.

Writing

• Accessibility is a reliability concern 2026-07-09

  Evaluation discipline and accessibility engineering are one posture applied to two substrates. The Reliability Layer's four runtime concerns already cover accessibility, once you notice that an interface is a system output and a user session is runtime.

• AI agents that test accessibility: two skills, released 2026-07-08

  Two Claude Code skills for agent-driven WCAG testing, released as open source with their test-contracts intact. What agent testing catches, what it cannot, and why the must-not-contain list is the piece worth defending.

• Trust and trustworthiness 2026-05-21

  AI agents are not trustworthy out of the box, and putting a trust-inspiring face on an untrustworthy system is dangerous. The three barriers to earning user trust - overconfidence, security, and reliability - and why they have to be faced before the face goes on. Part one of two.

• The AI Reliability Layer 2026-04-29

  A four-concern runtime taxonomy (verification, calibration, adversarial robustness, recovery) and a positioned framing of the territory it covers. Engages with the close adjacencies (AI TRiSM, NIST AI RMF, AI SRE).

• Veridi methodology: what the numbers actually mean 2026-04-26

  A deep-dive on the Veridi verification system and the Pragma + Praxis assessment frameworks that run on top of it. Calibration (selective Brier 0.0253, coverage 89%, abstention correctness 11/11, accuracy 99.5%), architecture, gaming countermeasures, limitations, and refinement backlog, grounded in published research.

Prototyping

Artifact-status view of the systems behind the work narrative. Each entry below shows what was shipped, what is in beta, and what is research-stage. Calibration internals and methodology versioning live in the Veridi methodology writeup; this page is the inventory.

In production

• NetterTech Events suite (nettertech.com). Five-plugin commercial WordPress event-management codebase. Free core published on WordPress.org. Production customers: Celtic Junction Arts Center, Ireland Network Minnesota. Surface area: RFC 5545 recurrence, custom database tables, AES-256 encrypted migration exports, REST API. Engineering envelope: 10,000+ automated tests; WCAG 2.2 AA across user surfaces; OWASP 2025 security posture; pre-commit and pre-push gates for static analysis, coding standards, and mutation analysis.
• Claude Code skills. Over 80 authored with declared test-contract blocks (verified 2026-07-26; see colophon for sourcing and exact counts). Coverage spans multi-agent team design, agent orchestration, skill-authoring quality frameworks, evaluation governance, methodology synthesis, plugin engineering, and accessibility. Test-contracts gate behavior on inspection rather than assertion. Two are published as open source (MIT) at claude-a11y-skills, test-contracts intact; further releases follow as they are scrubbed for portability.

Beta (invite-only)

• Veridi (veridi.org). Evidence-grounded fact-verification system run as a Canadian cooperative in formation. Methodology is published; the production service runs with admin-issued access codes while cost behavior under unrestricted load is validated. User-facing app supports English and French. For calibration data, architecture, gaming countermeasures, limitations, and refinement backlog, see the methodology writeup.
• Pragma + Praxis assessment frameworks. Run inside Veridi and also ship as standalone Claude Code skills. Pragma answers “what works?”; Praxis answers “what can you do?” Same epistemic discipline as Veridi: calibrated confidence, explicit limitations, abstention as a first-class output.

Research-stage

• Non-English source coverage for Veridi. Five language groups identified with candidate test claims (Chinese / Mandarin, Spanish / Portuguese, Arabic, Japanese / Korean, Hindi / Urdu). In-language verification specialists are not yet implemented; the methodology documents the gap and routes non-English claims through English-source paths with calibrated confidence reflecting that constraint. Path to ship sits behind the harder-validation-suite work in the methodology refinement backlog.
• Cross-vendor validation for Veridi. First 8-claim Gemini 2.5 Pro pass complete (50% verdict agreement with informative disagreements documented). Expansion to a 20-claim representative mix on the refinement backlog; restricted by frontier-model token spend.

Earlier shipped systems (linked to /work for context)

• Anaplan internal design system. 3+ years as sole continuous contributor; cross-functional adoption across product engineering and design. The composable abstraction discipline behind Veridi’s agent infrastructure and the Claude Code skills set was laid here.
• HIPAA-compliant systems at National Marrow Donor Program (now Be The Match). 7 years across internal and public surfaces; PGP-secured whistleblower system enabling anonymous misconduct reporting.

Coming

A newsletter for new writing on the AI Reliability Layer opens once the cornerstone series is two posts deeper than launch state. The RSS feed is the canonical subscribe surface in the meantime.