Claude Code and Codex generate. Agent Zero orchestrates. Paperclip delegates. Hermes serves. Faster than any single human or dev team — production code, not demos.
AI-assisted development with sharp engineering judgment. I know when AI output is correct, when it needs editing, and when to write manually. Speed without technical debt.
SHA-256 hash-chained audit trails on every decision. SOC 2-mapped. EU AI Act Article 12-ready. I don't claim compliance — I prove it.
Everything I build is shareable knowledge. My methodology, frameworks, and implementation patterns are designed to be replicated — not locked behind a proprietary wall.
Watchdog monitoring catches failures before users notice. Exponential backoff on transient errors. Deterministic failures alert me — they don't hallucinate a fix.
The implementation layer is what separates a demo from a deployed system. I move at the speed the climate crisis demands — because shipping velocity saves emissions.
Multi-industry AI concierge — WhatsApp, voice, web. Intent routing across 6 categories. Deterministic classification with specialized handlers. Production since 2025.
Built in 48 hours — SHA-256 audit trails, consent registry with instant withdrawal, mandatory HITL gates. Now inherited by every deployment. Production code, not a policy document.
Multi-agent stack (Hermes + Paperclip + Agent Zero) runs 40+ hours without intervention. Watchdog cron detects and recovers from failures before humans notice.
Legal, dental, hospitality, real estate, recruitment, business services. Same implementation layer, different surfaces. Framework is duplicatable — built once, deployed anywhere.
| What You Need | My Evidence |
|---|---|
| JavaScript, Node.js, React, Express | ✓ Node.js production stack. WebSocket gateways, REST APIs, React frontends across 6 industries. Portfolio: aithatbooks.co.za |
| PostgreSQL, Docker, Git | ✓ Postgres for guest memory + CRM data. Docker for Hermes, Paperclip, Agent Zero. Git daily across all projects. |
| Daily AI coding tool proficiency | ✓ Claude Code, Codex, Agent Zero, Paperclip, Hermes — in production, not just installed. Multi-agent orchestration means AI tools call AI tools. |
| Effective prompts & context for AI | ✓ 136+ skills with system prompts, custom instructions, and project rules. Every deployment inherits context documentation automatically. |
| Judgment: AI output vs manual code | ✓ Sharp distinction in my stack. Deterministic handlers for classification. AI for composition and generation. HITL gates for state changes. |
| Architect web application systems | ✓ Intent routing + graduated authority + multi-agent orchestration. The implementation layer — not just a React component. |
| 3+ years software development | ✓ 2+ years building production AI systems. Multi-agent orchestration, 6 industries, 40h+ autonomous runtime. Before that: 20 years in B2B tech. |
| 2 shipped products | ✓ Hermann Concierge (6 industries) + DentaLink AI (dental practice automation) + Compliance Framework (inherited by all deployments). |
| Python scripting | ✓ Python daily — data pipelines, InfraNodus API integration, audit trail generation, watchdog cron, deployment automation. |
| Open Source maintainership | ✓ Methodology and frameworks designed to be shared. Not a maintainer yet — but the ethos matches. See: github.com/Joche101/ai-portfolio |
| Dedication to fighting climate change | ✓ The implementation layer I build reduces waste — shipping faster with fewer resources. Same urgency, different crisis. Ready to direct that velocity at climate. |
Two years ago, I was writing every line of code by hand — debugging WebSocket protocols, hand-rolling HMAC auth, spending hours on what AI now does in minutes.
Today, AI is my default working surface. I prototype with it. I deploy with it. I monitor with it. Claude Code generates scaffold while I think about architecture. Codex catches edge cases I would have found at 2am in production. Agent Zero orchestrates tasks across my stack while I sleep.
But here's what most people miss: AI doesn't replace engineering judgment. It amplifies it. The difference between someone who "uses Cursor" and someone who operates an AI-native development stack is the implementation layer — knowing when AI output is correct, when it needs editing, and when to write code manually. That's the judgment I bring.
What changed for me is velocity. I shipped a compliance framework with SHA-256 audit trails in 48 hours — something that would have taken a team weeks. I built multi-agent orchestration that runs 40+ hours autonomously. Not because I'm faster than other engineers, but because I stopped treating AI as a tool and started treating it as an operating system.
Climate change is the defining challenge of our time. The tools exist. The data exists. What's missing is the implementation velocity to turn knowledge into deployed systems at the speed the crisis demands. That's what I do.
A dental client needed POPIA compliance for their AI concierge. I built an immutable audit trail system that now ships with every deployment — SHA-256 hash chaining, consent registry with instant withdrawal, mandatory HITL gates for PII operations.
Not a demo. Not a concept. Production code. Every AI decision traceable, every state change logged. SOC 2-mapped, EU AI Act Article 12-ready. Built once, reused everywhere.
Full portfolio and live demos:
John Bianchina · +27 78 914 0260 · john@aithatbooks.co.za
Johannesburg, South Africa · Fully Remote · GMT+2 (overlap with PST mornings)