M87 Studio
[Mac McFall / Engineering work]

Systems that
earn trust.

I design and implement AI-assisted systems where authority, evidence, and operational responsibility stay visible—from evaluation and agent governance to the application layer.

GitHub Request résumé
[What I bring]

Technical judgment that survives contact with production.

The through-line is not a particular framework. It is converting unclear authority and failure boundaries into systems a team can inspect, operate, and improve.

  1. 01

    AI systems and evaluation

    Design evaluation boundaries, authority gates, and evidence trails for AI-assisted systems that must remain reviewable.

  2. 02

    Platform engineering

    Turn ambiguous workflows into maintainable services with explicit interfaces, operational constraints, and recovery paths.

  3. 03

    Full-stack implementation

    Build with Python, TypeScript, Next.js, React, Postgres, Supabase, Docker, Kubernetes, and Vercel.

  4. 04

    Verification and security

    Treat tests, claims, secrets, access boundaries, deployment evidence, and reversibility as part of the product.

[Selected evidence]

Start with the artifact.

Each sample states what is verifiable now and what remains intentionally unclaimed. That distinction is part of the engineering work.

Spine Lite

Public source · MIT

A policy-enforcement hook for Claude Code that separates model proposals from authorized execution.

Evidence boundary: The public repository and MIT license are verified. The project documents its original seven-class decision taxonomy; broader runtime support is not claimed.Inspect source

Governed Swarm

Public source · BSL 1.1

Multi-agent coordination organized around explicit authority, gates, immutable evidence, and reversible execution.

Evidence boundary: The canonical repository and BSL 1.1 license are verified. Release, deployment, and support maturity remain under evidence review.Inspect source

Spectrum

Private experiment

Audio-analysis engineering spanning multi-resolution spectral analysis, loudness measurement, and browser-based signal processing.

Evidence boundary: The work is intentionally presented as an experiment, not a production promise. Technical discussion and samples remain subject to rights and confidentiality review.See current classification
[Employment boundary]

A practice—not a competing startup.

M87 Studio is my independent practice: the place I maintain open-source work, develop portfolio evidence, and keep my delivery discipline sharp. It is not a venture I intend to operate in conflict with an employer.

For the right full-time role, the role is the priority. Outside work remains bounded by the employment agreement, confidentiality obligations, and a clear prior-inventions disclosure.

[For hiring teams]

Bring the hard system question.

I can provide a résumé, walk through the engineering decisions behind these artifacts, and discuss where this discipline fits your team.

Start a conversationView GitHub