
Joshua E
AI Agentic Developer, Multi Agent System, Jarvis AI, Relevance AI
Skills

Bekijk mijn diensten


Werkervaring
OpenClaw & Hermes Agent Specialist
Hunab • ZZP
Jan 2026 - Present • 8 mos
I set up, secure and extend self-hosted AI agents for founders, small businesses and teams, using OpenClaw and Hermes Agent. My work covers installation on VPS servers, Mac mini, Windows (WSL2) and Docker; connecting WhatsApp, Telegram, Slack and Discord; configuring models such as Claude, GPT and local Ollama models with fallbacks; and building custom skills around each client's workflows. Security is part of every setup: private access through Tailscale, strong gateway authentication, sender pairing, sandboxing, command approvals and a security audit before handover. Each project ends with a working agent, a written runbook and a handover call, with optional monthly care plans for updates and monitoring.
AI Automation Developer (n8n, MCP & Integrations)
Fiverr • Freelance
Nov 2024 - Present • 1 yr 10 mos
I connect AI agents to the tools businesses already use. I build n8n workflows that agents can call as tools, set up MCP servers for GitHub, Notion, Google Drive and databases, and create custom MCP servers for APIs that don't have one. Typical projects include email triage with draft replies for approval, morning briefings on Telegram or WhatsApp, lead intake into a CRM, receipts from WhatsApp photos into Google Sheets, and automated weekly reports. I design every automation with approval steps and least-privilege access, so agents can act without being able to cause damage, and I document each workflow so clients can maintain it themselves.
AI Agent Security & Deployment Engineer
Gutab • Parttime
Apr 2026 - Aug 2026 • 4 mos
I audit, harden and maintain self-hosted AI agent deployments. My work includes running security audits and fixing what they find: gateways exposed to the internet, weak tokens, open direct-message policies and Docker ports that bypass the firewall. I move API keys into secure secret storage and set up sandboxing and command allowlists. I use reader-agent designs to limit prompt injection from emails and web pages. Before installing any community skill or plugin, I review its code. I also set up backups with tested restores, uptime monitoring and a safe update routine on the extended-stable release line. Clients receive a written report explaining every change, so they understand how their agent is protected.