Resume
Jeffrey Michael Johnson
AI engineer working on agent orchestration and voice-first agent systems: control planes, guardrails, and evaluation discipline. Builder of GBuild, a notarized macOS AI build environment driven by voice; also shipped a 6-platform app live on the App Store with a hand-designed UI/UX and motion system. Prior career in institutional finance and venture-backed startups.
Relevant Experience
Founder & Principal Engineer · GBuild (codename Mother) · 87n1
2026 – PresentRust (tokio) · TypeScript · Electron · React · Cloudflare Workers/DO/Queues/KV/R2 · Python (ML + voice) · Supabase · Stripe · Telnyx · LiveKit
- Built a notarized, auto-updating macOS AI build environment: 633,438 hand-written LOC across a Rust sidecar (916 registered JSON-RPC methods, panic-isolated dispatch, 810 source files) and an Electron/React front end, with 275 notarized releases distributed via Cloudflare R2 in the first 81 days
- Designed a 14-Worker (in repo) Cloudflare edge backend (Service Bindings, Durable Objects with SQLite + Hibernation, Queues, KV, R2, Analytics Engine, Cron Triggers) covering auth, telemetry, an autonomy engine, a voice/telephony hub, and a deploy-gated billing rail
- Implemented identity from primitives, no SDKs: magic-link auth, 4-provider OAuth2 with hand-rolled PKCE, a 16-connector token vault with AES-GCM encryption at rest, and dual-token Supabase JWT minting for RLS-scoped Realtime
- Built ThingStore, a signed app marketplace: Ed25519 publisher signing over a SHA-256 content-hash chain, a 5-layer automated publish gate (schema validation, secret/PII scan, capability-ceiling linting, CSP static analysis, LLM risk review), egress-allowlist scrutiny, and a revocation feed that holds offline
- Built The Foundry, an on-device model line: fine-tuned ModernBERT/DeBERTa classifiers on synthetic data generated by typed-field corruption, evaluated against human-labeled gold sets behind paired-bootstrap 95% LCB significance gates with a champion/challenger harness, and shipped a Rust runtime ranker with a versioned feature contract to the Python trainer
- Engineered agent-runtime infrastructure: a pluggable AgentDriver trait behind 5 first-party drivers plus sandboxed bring-your-own-agent registration, coordination groups over a pub/sub bus, advisory file leases with deadlock-avoiding sorted acquisition, and a policy-as-code governor (task class → model tier, cost ceiling, tool least-privilege) enforced by a build-failing conformance test
- Cut per-turn LLM cost by trimming ~96 MCP tool schemas to each agent’s capability set, byte-deterministically so the prompt cache still hits
- Hardened the untrusted-code surface: macOS Seatbelt (sandbox-exec) SBPL profile generation, per-app Electron session partitioning with deny-by-default permission handlers, and a loopback HTTP CONNECT egress broker that defends against DNS rebinding and SSRF
- Built AI-safety operational infrastructure across the agent fleet: off-switches, circuit breakers, cooldowns, an append-only audit trail, and absence-based alerting that treats a missing heartbeat as the signal
OpenAI Realtime · Gemini Live (3.1 Flash Live) · MLX Whisper · Kokoro TTS · Telnyx · TypeScript · Rust
- Architected a voice-first control plane spanning dictation, conversational, and realtime modes (OpenAI Realtime, Gemini Live) with on-device MLX Whisper STT and Kokoro TTS: an intent classifier and router turns speech into tool calls that operate macOS apps, files, and folders, GBuild itself, and the agent fleet, guarded by destructive-intent confirmation and transcript redaction
- Built a custom Voice Harness Lab plus training center for voice agents: real multi-turn conversations against live harnesses where per-turn human feedback becomes an approved patch to the harness source effective on the next run, alongside scripted voice skits, an intent-router bench, consent-gated dataset export, and replayable session journals
- Designed a proactive voice agent pattern (the agent probes for ideas, interests, and goals instead of waiting for prompts) and demonstrated it in live internal sessions completing a PRD and personal and business CRM records in real time from the running transcript
- Programmed AI agents to place real phone calls over Telnyx telephony behind a Cloudflare Workers comms hub; deployed internally in a test environment for internal testers
Independent Software Engineer
2025 – PresentFigma · React Native · Reanimated · Skia · Expo · Supabase · Mapbox · SwiftUI (watchOS) · LiveKit
- Designed the "Ethereal Glass" UI/UX and animation system in Figma (color system, typography, glassmorphism tokens, 8 category-specific glow colors, 6-phase prayer connection choreography), then translated the wireframes and prototypes into a production React web app before pivoting to React Native based on design-fidelity requirements
- Authored a 470-line animation preservation spec mapping every Framer Motion interaction to Reanimated and Skia equivalents; iterated on the front-end implementation through a dedicated animation lab (prayerConnections, memorialHub experiments) to hit a 60fps native target
- Shipped on the App Store across 6 platforms (iOS, Android, web, watchOS, iPadOS, macOS) with 46 edge functions, 198 database migrations, native watchOS companion, and real-time subscriptions
- Built 6-dashboard admin command center with 3-tier AI behavior monitoring, Datadog RUM, and PostHog analytics
Node.js (ESM) · Claude Code CLI · MCP Protocol · ChromaDB · Docker
- The Smart MCP Loader + 13 skills + 7 specialized agents that form the operating system I build inside. NLP-scored prompt routing across 11 adaptive tiers, MIT License.
Node.js (ESM) · Claude Code Hooks · Regex NLP Scoring · node:test
- 183 passing tests, 0 failures. Classifies prompts on a 4x4 complexity/stakes matrix, routing to 11 adaptive tiers. Adversarial-tested with multi-agent parallel validation.
Python · LangGraph · FastAPI · Pinecone · MCP Protocol
- 7 specialized LangGraph agent types, 179 commits of WordPress hook-mapping research. Pinecone vector retrieval with hybrid knowledge graph architecture.
Co-Founder · FourPlay Football (Tech Startup)
2014 – 2020- Co-invented and patented (US9569922B1) a mobile sports platform for iOS and Android
- Led product from concept through seed round and Series A funding, managing cross-functional delivery across design, engineering, and marketing
Owner & Operator · Get Some Fitness Studio
2017 – 2024- Built and operated a fitness business for 7 years: full P&L responsibility, client acquisition, staffing, and day-to-day operations
Senior Portfolio Manager · Morgan Stanley
2011 – 2015- Managed multi-million-dollar client portfolios with quantitative risk analysis, executive-level stakeholder communication, and regulatory compliance
Financial Advisor · Raymond James & Associates
2008 – 2011Skills
Languages
TypeScript (strict), Rust (async/tokio, production), Python, Swift, SQL (Postgres + SQLite), C, Objective-C++, Bash, HTML, CSS
Design & Front-End Craft
Figma, UI/UX Design, Design Systems, Wireframing, Prototyping, Motion Design, Interaction Design, Design-to-Code, Framer Motion, React Native Reanimated, React Native Skia, After Effects, Glassmorphism, Typography, Responsive Design, Accessibility
Libraries & Frameworks
React, Next.js, Node.js, Electron, React Native, Expo, React Three Fiber / Three.js, FastAPI, SwiftUI, LangGraph, Tailwind CSS
AI & Agents
Claude API, Claude Code Hooks, MCP Server Authorship & Lifecycle Management, Multi-Agent Orchestration, Pluggable Agent-Driver Architecture, Policy-as-Code Agent Governance, Prompt-Injection Defense, LLM FinOps / Cost Ceilings, Agent Evaluation Harnesses, Eval Design / LLM-as-Judge, RAG with Reranking, Model & Cost Routing, Agent Observability, Prompt Engineering, Guardrails & Tool-Gating
Voice AI
Realtime Voice Agents (OpenAI Realtime, Gemini Live), On-Device STT/TTS (MLX Whisper, Kokoro), Dictation / Conversational / Realtime Modes, Voice Intent Routing & Classification, Voice Tool Calling (OS, App, and Agent Control), Proactive Agent Design, Voice Harness Labs & Agent Training, Telephony Agents (Telnyx), Destructive-Intent Guards, Transcript Redaction
ML Engineering
Transformer Fine-Tuning (ModernBERT/DeBERTa), Synthetic Data Generation (Typed-Field Corruption), Gold-Set Evaluation, Paired-Bootstrap Significance Gates, Champion/Challenger Promotion, On-Device Model Runtimes (MLX, Ollama, candle), Feature-Contract Versioning (Python Trainer ↔ Rust Runtime)
Edge & Cloud
Cloudflare Workers, Durable Objects, Queues, KV, R2, Service Bindings, Analytics Engine, Zero Trust Access, Cron Triggers, Vercel, AWS, Supabase, Docker, GitHub Actions
Security
Supply-Chain Signing (Ed25519 + SHA-256 Content Addressing), Secret Scanning & Redaction, macOS Seatbelt Sandboxing, SSRF / DNS-Rebinding Defense, CSP Static Analysis, AES-GCM Envelope Encryption, OAuth2 + PKCE from Primitives, RLS-Preserving Vector Search
Native macOS
Electron Hardening (Session Partitioning, IPC Allowlisting), N-API Addons (Objective-C++), Relocatable CPython Embedding, Notarization & Mach-O Nested Signing, TCC / Entitlements, Keychain, launchd
Architecture & APIs
REST API Design, JSON-RPC at Scale, Event-Driven Architecture, WebSockets, Durable Workflow Execution (DAG / Topological Sort), Distributed Locking (TTL Leases), Serverless Functions, Webhooks, Real-Time Subscriptions, Agile / Iterative Delivery
Practices
Contract-vs-Implementation Drift Gates, Ratchet Baselines, Supply-Chain CI (gitleaks, cargo-deny, License Gating), Local/CI Gate Parity, Architecture Decision Records, Vitest, Playwright, Git, PostgreSQL, n8n
Selected Projects
AI-native macOS build environment orchestrating a fleet of coding agents behind 5 first-party drivers: Electron 42 + an 810-file Rust sidecar over a 916-method JSON-RPC surface, on-device voice (MLX Whisper, Kokoro TTS), local Qwen3 RAG, full MCP lifecycle management. 6,308 commits and 275 notarized auto-updating releases in its first 81 days.
A signed marketplace for AI-built apps: Ed25519 publisher signing over a SHA-256 content-hash chain, a 5-layer automated publish gate, a per-app loopback egress broker, and revocation that holds offline. Supply-chain security for an app store, built solo.
GBuild’s on-device model line: ModernBERT/DeBERTa classifiers fine-tuned on synthetic data from typed-field corruption, promoted only past paired-bootstrap 95% LCB significance gates on human-labeled gold sets. The 97.5%→62% overfit its gold set caught is published, not hidden.
The Smart MCP Loader + 13 skills + 7 specialized agents that form the operating system I build inside. MIT License.
183 passing tests, 0 failures. Zero-dependency prompt scoring engine classifying prompts on a 4x4 complexity/stakes matrix, routing to 11 adaptive tiers.
Education
AWS Cloud & AI Training
6 Skill Builder foundation certificates (Cloud Practitioner, Gen AI, ML) plus the official Generative AI Developer Professional (AIP-C01) exam-prep track across all 5 domains; Practice assessments scored 100 on Domains 1, 2, and 5 (68% of scored exam content). Proctored exam pending
University of Mississippi
Finance · 2 Years