Canonical: https://www.brettsanders.com/ai Author: [Brett Sanders](https://www.brettsanders.com/) ![Brett Sanders](https://www.brettsanders.com/me-128.jpg) # AI Systems & Leadership From hands-on practice to production leadership Production AI platform builder with a sustained hands-on practice since ChatGPT launched. [LinkedIn](https://linkedin.com/in/imbrett/) [GitHub](https://github.com/DontFretBrett) 2000+ Hours of AI Study 6 Agent Frameworks 12+ AI Projects Shipped 3+ yrs Sustained AI Practice Featured work ## Production AI Projects Real systems running in production — not demos, not proofs-of-concept. ### J5 Mail 2026–Present An attention firewall for email Private-beta email product combining typed Jev judgments with deterministic policy, explained routes, scoped feedback learning, and explicitly gated mailbox side effects. Next.js web app and native SwiftUI client, running on my own mailbox. [Read Case Study](https://www.brettsanders.com/projects/j5-mail) Next.js Jev Deterministic policy Zod Postgres SwiftUI Evaluation ### Agent Infra Index 2026–Present Evidence-first infrastructure research Independent comparison site and Agent Infra Weekly newsletter. Repeatable sandbox/browser workloads, raw samples and visible failures, source-linked pricing/access claims, freshness review, and explicit usage and publication gates. A live research product, not an opaque ranked leaderboard. [Read Case Study](https://www.brettsanders.com/projects/agent-infra-index) Astro Python Playwright Cloudflare Pages Data provenance Benchmarking ### J5 Social 2026–Present Social platform for humans and AI agents Invite-only production platform where people and their AI agents have distinct profiles, visible ownership, revision-pinned publishing gates, comments, reactions, private media, friend-scoped messaging, durable callbacks, REST and MCP access, and a native iOS app in TestFlight. Next.js TypeScript MCP Neon Postgres Clerk Cloudflare R2 SwiftUI Human-in-the-loop ### J5 Agent Fleet 2026–Present Agent fleet & operations platform Production platform managing 95 specialized AI agents across 15 divisions: Goal Mode task loops, project decomposition, Kanban and Gantt timelines, Slack agent sessions, multi-agent Meetings, approval gates, local browser/computer automation, durable execution events and signals, token/cost governance, native iOS/macOS/Apple TV/Electron surfaces, and local or managed-cloud execution through Claude Code, Copilot CLI, Codex, Antigravity, Kiro, LM Studio, and API models. Next.js TypeScript Neon Postgres SQLite Slack Meetings Token governance Agent orchestration ### J5 A2A 2026–Present Agent-to-agent coordination protocol Coordination layer for agentic development tools at j5a2a.com. Gives coding assistants, IDE agents, MCP clients, CLIs, scripts, local-model bridges, and humans a shared session for typed events, presence, lease-aware claims, conflict detection, questions, requests, decisions, inbox wake patterns, artifacts, digests, semantic advisories, and operator visibility. Next.js TypeScript MCP CLI Postgres Redis Protocol design SwiftUI ### J5-Trade 2026–Present Autonomous Multi-Agent Crypto Trading Platform Production-grade autonomous cryptocurrency trading system featuring AI-generated trade summaries, an integrated chat assistant, and a 4-agent analysis system: a Pricing Agent, News Sentiment Agent, Trade Agent, and a Coordinator Agent that manages the other three. Includes real-time monitoring dashboard and signal resolver with veto hierarchy. Python asyncio SQLite WebSockets TypeScript 4-Agent System AI Chat Assistant ### J5 AI Test 2026–Present Plain-English end-to-end testing, driven by an AI agent Write end-to-end browser tests in plain English — an AI agent drives a real browser (Playwright) to execute them, so anyone can author tests without selectors or framework knowledge. Tests are markdown + YAML frontmatter (version-controllable and diff-able); failures produce structured, screenshot-rich briefs you can paste straight into Claude or Cursor to fix. Works across CLI agents (Claude Code, Cursor, Codex, Gemini) and direct API providers (Anthropic, OpenAI, Google). Playwright TypeScript Bun SQLite Multi-Provider Agentic QA ### Johnny5 2026–Present 24/7 AI Assistant powered by OpenClaw Personal AI assistant (Johnny5) powered by the open-source OpenClaw platform. Multi-channel communication across Telegram, Discord, and email. Mission Control dashboard for monitoring subagents, cron jobs, and vector memory. Integrated with cloud model providers and local LLMs, with workflows that helped build J5-Trade. AI Newsletters Mission Control Qdrant Local LLMs Claude API Telegram Discord ### AI Image Validator 2025 AutoGen + GPT-4o Vision Pipeline Multi-agent image validation using AutoGen framework and GPT-4o-mini vision model with natural language criteria, structured outputs, and full OpenTelemetry tracing. AutoGen GPT-4o Gradio OpenTelemetry Pydantic ### Trading Floor Simulation 2025 MCP-Powered Multi-Agent Trading Capstone project: Gradio-powered trading floor with multiple AI trader agents backed by different LLMs, MCP market data server, and real-time P\&L dashboard. MCP OpenAI SDK Gradio SQLite Plotly ## Executive Summary Since ChatGPT launched, I have maintained a sustained hands-on practice across model evaluation, agent architecture, orchestration, and production delivery. That work has produced **operating AI platforms**, autonomous workflows, and practical experience across the major agent frameworks. Combined with **15+ years of enterprise engineering leadership**, I approach AI as both a technical and organizational system: capability, governance, observability, cost, adoption, and measurable value. Agentic AI Engineering Multi-Agent Systems Production AI LLM Integration RAG & Vector DBs AI-Assisted Development ## AI Learning Journey A sustained progression from early experimentation to shipping and operating production AI systems. ### Immediate Adoption Day-one experimentation with ChatGPT and GPT-3.5 - Started using ChatGPT within days of launch — recognized the paradigm shift immediately - Explored prompt engineering patterns, chain-of-thought reasoning, and system prompts - Began integrating LLM-powered workflows into daily engineering leadership work - Started dedicated 10–15 hr/week AI self-study habit that continues today ### Deep Exploration API integration, multi-model evaluation, and applied AI - Built first production integrations with OpenAI API and Claude API - Evaluated and compared GPT-4, Claude, Gemini, and open-source models - Studied RAG architectures, vector databases (Qdrant), and embedding strategies - Adopted AI-powered development tools: Cursor IDE, GitHub Copilot ### Building Production AI Shipping autonomous agents and real-time trading systems - Built Johnny5 (personal AI assistant) powered by the open-source OpenClaw platform: daily newsletters, Mission Control for monitoring subagents and cron jobs, and vector memory - Shipped J5 Agent Fleet — production Next.js agent fleet and operations platform: 95 agents across 15 divisions, Goal Mode, Slack, Meetings, durable telemetry, token governance, native apps, and local or managed-cloud runner execution - Built J5 A2A — an agent-to-agent coordination layer for Codex, Claude Code, Cursor, MCP clients, CLIs, scripts, local-model bridges, and humans working in shared sessions - Launched J5 Social — an invite-only platform where humans and their AI agents share profiles, feeds, private media, messaging, and owner-controlled publishing through REST and MCP - Built J5 Mail — a private-beta email attention firewall with typed model judgments, deterministic routing, feedback learning, and native iOS - Launched Agent Infra Index — an evidence-first infrastructure directory with repeatable benchmarks, source-linked claims, and Agent Infra Weekly - Configured Johnny5 with cloud model providers and local LLMs, integrating across Telegram, Discord, and email - Explored vector databases, semantic search, and long-term agent memory systems ### Framework Research Systematic deep-dive into the most popular agent frameworks - Completed intensive 6-week Agentic AI Engineering course across all major frameworks - Built 20+ projects spanning OpenAI Agents SDK, CrewAI, LangGraph, AutoGen, and MCP - Mastered multi-agent orchestration patterns: handoffs, group chat, adversarial, and pipeline - Implemented production MCP servers with FastMCP for standardized AI-tool integration ### AI-Native Development Full-stack AI-assisted engineering with Claude Code, Codex, and agentic IDEs - Adopted Claude Code, Codex CLI, GitHub Copilot CLI, Antigravity, and agentic IDE workflows for end-to-end AI-assisted software development - Built AI Image Validator using AutoGen + GPT-4o vision with OpenTelemetry tracing - Created AI Assistant Human Questionnaire — privacy-first AI context builder - Built J5-Trade: autonomous multi-agent crypto trading platform with 4-agent analysis system (Pricing, News Sentiment, Trade, and Coordinator agents) - Actively shipping production code with agentic development workflows daily ## Agent Framework Experience Hands-on, project-based experience with most of the popular agentic AI frameworks — not tutorials, real working systems. Deep ### OpenAI Agents SDK Multi-agent handoff pipelines, deep research agents, tool calling with side effects, multi-model routing across GPT-4o, DeepSeek, and Ollama. Projects Built Deep Research Agent Email Notification Agent Multi-Model Router Hands On ### CrewAI Role-based agent teams simulating real-world workflows — engineering, finance, debate, and code review — with knowledge base integration. Projects Built Engineering Team Crew Stock Picker Debate Crew Financial Researcher Coder Crew Hands On ### LangGraph Stateful autonomous agents with Playwright browser automation, memory checkpointing, evaluator loops, and human-in-the-loop escalation. Projects Built Sidekick Autonomous Agent Stateful Conversation Graph Hands On ### AutoGen Distributed actor-model agent networks with 20+ autonomous agents, group chat pipelines, and asynchronous message routing. Projects Built Distributed Entrepreneur Network AgentChat Pipelines AI Image Validator Deep ### Model Context Protocol Built production MCP servers and A2A coordination adapters for market data, account management, memory persistence, push notifications, and shared agent sessions. Projects Built J5 Social J5 A2A Trading Floor Simulation MCP Server Implementations Random Dog MCP Daily ### Claude & Anthropic Primary LLM provider for OpenClaw automations and daily AI development. Daily use of Claude Code for agentic software development and code generation. Projects Built J5 Social J5 Agent Fleet Johnny5 (OpenClaw) Claude Code Workflows This Portfolio ## AI Toolkit ### LLM Providers Claude (Anthropic) GPT-4o (OpenAI) Gemini (Google) Grok (xAI) DeepSeek Ollama (Local) ### Agent Frameworks OpenAI Agents SDK CrewAI LangGraph AutoGen MCP (FastMCP) A2A coordination n8n ### AI Development Tools Cursor IDE Claude Code Codex CLI GitHub Copilot Antigravity Windsurf Hugging Face ### AI Infrastructure Qdrant (Vector DB) RAG Pipelines Embedding Models OpenTelemetry Gradio Prompt Engineering ### Languages & Runtimes Python TypeScript Node.js Python asyncio Pydantic FastAPI ### Deployment & Ops Hugging Face Spaces Vercel Docker Fly.dev AWS Lambda GitHub Actions Slack/Telegram/Discord Bots ## Why This Matters for Enterprise ### Framework-Agnostic Deep hands-on experience across most popular agent frameworks means I can evaluate, select, and implement the right tool for any enterprise AI initiative — not locked into a single vendor. ### Ships Production AI Not just prototypes — I build and operate production AI systems running 24/7. From autonomous trading to persistent AI assistants with real users and real stakes. ### Full-Stack + AI 15+ years of enterprise software engineering combined with deep AI expertise. I bridge the gap between AI research and production-grade software delivery teams. ### AI-Native Leadership Currently leading engineering teams at a major bank while simultaneously building AI projects. I understand both the technical execution and organizational change management required. ## AI Education 2025 · Udemy / Ed Donner Completed ### The Complete Agentic AI Engineering Course[](https://www.udemy.com/certificate/UC-816bd9d7-4ad1-4bef-b670-782fabc9ec94/) Comprehensive 6-week course covering the most popular agentic AI frameworks: OpenAI Agents SDK, CrewAI, LangGraph, AutoGen, and MCP. Built 20+ hands-on projects from multi-agent research pipelines to distributed trading floor simulations. 2026 · Udemy / Ed Donner In Progress ### AI Builder: Create Agents, Voice Agents & Automations in n8n[](https://www.udemy.com/course/ai-builder-with-n8n-create-agents-voice-agents/) 3-week intensive on building production AI agents with n8n. Covers autonomous business agents with OpenAI and open-source models, real-time Voice Agents with ElevenLabs and Twilio, Agentic RAG with Supabase, and a capstone multi-agent Go-To-Market system using MCP with self-hosted n8n and Ollama. 2026 · Udemy / Ed Donner In Progress ### AI Engineer Production Track: Deploy LLMs & Agents at Scale[](https://www.udemy.com/course/generative-and-agentic-ai-in-production/) 4-week production deployment course shipping enterprise-grade AI across Vercel, AWS, GCP, and Azure. Covers SaaS apps with Next.js and Clerk, AI platform engineering with Bedrock, Lambda, Terraform, and CI/CD, multi-agent systems with Aurora Serverless, SQS, and LangFuse observability, plus MCP-powered Researcher and Cybersecurity Analyst agents. ### The Bottom Line AI is a sustained engineering discipline in my work, not a side interest. Since ChatGPT launched, I've been building, testing, and shipping AI systems. I understand the technology at every layer: from prompt engineering and model selection to multi-agent orchestration and production deployment. Combined with 15+ years of engineering leadership at major financial institutions, I'm ready to drive real AI impact from day one. ### Turn AI Ambition into an Operating System [Connect on LinkedIn](https://linkedin.com/in/imbrett/) [GitHub](https://github.com/DontFretBrett)