Hey, I'm 👋
Griffin Long
I don't out-type teams — I build and direct the AI agent tooling that lets one person ship at team scale.
My story
How a line cook ended up building AI agent systems.
I didn't take the traditional route into software. I taught myself to build with AI while working as a line cook and taking college classes — experimenting with ChatGPT on the side until it turned into something real.
In March 2025 I turned that into GRAIsol. In the ~15 months since, it's grown into roughly 10,800 commits across 100+ repos: an open-source agent runtime (agent-afk), a 369K-LOC AI GTM-automation platform, and 9 published open-source packages.
The throughline: I don't out-type teams — I build and direct the AI agent tooling that lets one person ship at team scale. Same energy I brought to the kitchen — high tempo, reliable under pressure — now pointed at shipping software end to end.
Things I’ve built
Open-source tools, production platforms, and the agent infrastructure behind them.
agent-afk
agentafk.com ↗open-source TypeScript agentic coding-agent runtime/harness
LOC
~308K TS/TSX
npm versions
408
commits
1,542
license
Apache-2.0
36-instance non-networked subset, own-harness eval, web tools disabled
AgentGRAI
Modular Next.js / React AI lead-intelligence and outbound platform with multi-stage enrichment, async queue architecture, Gmail outbound engine, and AI SDR layer
LOC
~369K
commits
~1,900
timeframe
~5 months
7 MCP Servers
ElevenLabs Voice MCP
tools: 23
ClickUp MCP
tools: 19
Cursor Agent MCP
merged PRs: 40
E2B MCP
Twilio MCP
Smartlead MCP
GPT Image MCP
agent-framework
Claude Code / agent-afk plugin for autonomous task execution through composable skills, subagents, and hooks
gadscli
Async Rust CLI for the Google Ads API, published to npm (npm-only)
A/B Experimentation Platform
5-service Docker Compose system with chi-squared significance from scratch, SHA-256 user bucketing, and database-level race-condition handling
On-Device AI Tools
Local macOS menu-bar dictation (mlx-whisper) and system-wide inline ghost-text completion (MLX Swift) for Apple Silicon
Tools I build with
The stack behind the projects above.
Languages
Frameworks & Runtimes
AI & Agents
Data & Infrastructure
Experience
From founding GRAIsol to embedded contract work and independent consulting.
Founder and AI Systems Engineer
GRAIsol · Daytona Beach, FL
- Built and open-sourced agent-afk (agentafk.com), a TypeScript agentic coding-agent runtime/harness: ~308K LOC, 1,542 commits, 408 published npm versions, Apache-2.0.
- Architected AgentGRAI, a modular Next.js / React AI lead-intelligence and outbound platform (~369K LOC, ~1,900 commits in ~5 months): multi-stage enrichment on async QStash queues with Redis-semaphore concurrency control.
- Extended it with a Gmail outbound engine (multi-inbox rotation, AI sequences, open/click tracking, AI reply classification) and an AI SDR layer with ICP scoring and human-handoff routing.
- Published 9 open-source packages across npm and GitHub: 7 MCP servers (ElevenLabs Voice – 23 tools, ClickUp – 19 tools, Cursor Agent – 40 merged PRs / SSE on Vercel, E2B, Twilio, Smartlead, GPT Image), plus agent-afk and gadscli.
- Ran a one-person AI-leveraged studio delivering 30+ client and product builds.
- Accumulated ~10,800 GitHub commit contributions, 193 PRs, and 107 repos since Mar 2025.
AI Systems Engineer (Contract)
Atlas Digital · Remote
- Embedded directly with client leadership to design and deploy an end-to-end CRM and agent automation pipeline: lead intake, enrichment, scoring, qualification, and outbound handoff in production.
- Built customer-facing agent workflows on Claude Agent SDK and OpenAI with structured fallback/retry logic and production observability; owned architecture end-to-end from design through deployment.
Independent AI Consultant
Remote · Remote
- Fuly Kids (fulykids.com): Diagnosed recurring LLM over-spend in FulyScan (AI food-scanning nutrition tool); shipped fix via merged PR adding Supabase/Postgres caching so repeat scans reuse stored results instead of re-invoking Claude.
- adside.ai: Delivered prioritized UX and product audit covering onboarding friction, cognitive overload, missing agency-facing tooling, data-freshness gaps, agent latency, and strategic positioning.
Let’s talk
Open to agentic AI engineering, technical leadership, and contract or consulting work — the fastest way to reach me is email.