Claude Code Coach & Agentic AI Trainer — Teach Us to Build, Help Us Sequence Real Projects

I'm the founder of a 26-year commercial real estate brokerage spinning out an AI-driven intelligence platform. I'm bringing on a fractional CTO and a senior builder in parallel — but what I'm missing is a teacher who has deep, hands-on production experience with Claude Code, Anthropic's MCP (Model Context Protocol), and modern agentic AI workflows, and who genuinely enjoys transferring that capability to others. This role is part technical coach, part trainer, part embedded advisor. You will not be the primary builder of production code. You will be making sure my team becomes self-sufficient using these tools, while occasionally building reference implementations and coaching the senior builder on Claude Code patterns specifically. Who this engagement serves: - Me (the founder, non-technical) — I want to understand what these tools can and can't do well enough to make smart product decisions. - My senior builder — They'll write production code but a coach with deeper Claude Code and MCP expertise accelerates their ramp. - My operations team — One VA and one or two contractors who currently work in GoHighLevel, N8N, and Airtable. I want them gradually upskilled into doing AI work they currently can't. - My fractional CTO — A peer to consult on Claude-specific tool selection and architecture decisions. The concrete project universe: I have a documented list of seventeen automation categories I want built across my brokerage operation. Some are pure Claude Code work, some are N8N workflows that call the Claude API, some are document generation tasks, and some belong in GoHighLevel automation. Part of your job is helping us sequence and triage the list — which projects are right for Claude Code specifically, which belong elsewhere, and what order to build them in. Representative projects from the list — these are illustrative, not exhaustive: - BOV and Offering Memorandum generation. Currently 2-4 hours of manual work per deal. We want a workflow where structured property data plus comp research generates a draft document in our template, with human review on the way out. - Post-call transcript analysis pipeline. Every CallRail recording processed through Claude to extract motivation signals, psychographic profile, urgency score, and suggested next action. Outputs pushed to our CRM as structured fields. - Motivation scoring engine. A scoring algorithm that combines call sentiment, SMS responses, newsletter engagement, ownership tenure, and psychographic classification into a 0-100 score per owner. Updates on every new data event. - Buyer-seller matching workflow. When a seller crosses a motivation threshold, the system queries the buyer database for matches, ranks the top three, and sends the agent a context package. - Obituary and probate monitoring agent. A scheduled agent that monitors obituary feeds, fuzzy-matches against our owner database, and creates high-priority tasks when matches cross a confidence threshold. - Deal structure analysis tool. Paste a lease or term sheet and Claude analyzes it against standard market terms, flagging issues, unusual provisions, and negotiating opportunities. - Agent briefing package generation. When a new lead is assigned, the system generates a briefing document with motivation context, transcript summary, psychographic profile, and suggested conversation approach. Your job is to look at the full list (which I'll share with serious candidates), tell us honestly what belongs in Claude Code versus what belongs elsewhere, teach us how to build the Claude Code pieces ourselves with your guidance, and build the trickiest reference implementations so we can study the pattern and extend it. What this person will do, concretely: - Run weekly 1-2 hour live sessions teaching us how to use Claude Code, MCP servers, agent frameworks, and related tools effectively. Recorded for replay. - Review the full automation list with us and produce a written triage document: which tool for which job, what to build first, what to defer, what we're missing. - Build small reference implementations we can study, modify, and extend. Not production code at scale — teaching code that demonstrates the right pattern. - Coach the senior builder on Claude Code idioms, MCP server design, and agentic workflow patterns through code review and pairing sessions. - Stay current on what's shipping in the Anthropic ecosystem and tell us what actually matters versus what's hype. - Answer ad-hoc questions through Slack or async video. - Build internal documentation that lets new team members ramp without re-asking the same questions. What this role is NOT: - This is not a builder role. If you want to be paid to ship our entire production codebase, this is the wrong post. - This is not a fractional CTO role. Architecture and product decisions belong to my fractional CTO. - This is not a one-off training engagement. I want a 3 to 6 month relationship. - This is not a generic "AI consultant" role. I want depth in the Anthropic ecosystem specifically. Required experience: - Substantial production experience with Claude (Anthropic's API), Claude Code specifically, and MCP servers. Months or years, not weeks. - Experience teaching, training, mentoring, or transferring technical skills to non-experts. - Comfort explaining technical concepts to non-technical people without being condescending or oversimplifying. - Working knowledge of the broader modern AI stack: OpenAI, vector databases, RAG patterns, agentic frameworks, workflow orchestration tools. - Strong written communication — much of the teaching will happen async through documentation and recorded explanations. Strong plus: - Public teaching footprint: YouTube channel, blog, GitHub repos with educational content, conference talks, course authorship. - Experience teaching teams that include non-developers (operations people, analysts, founders) alongside developers. - Familiarity with GoHighLevel and N8N (so you can speak intelligently about when to use Claude Code versus when to use those tools). Engagement structure: - 5-15 hours/week, hourly through Upwork tracker. - Weekly live teaching session (1-2 hours) at a fixed time we agree on. - Plus async availability for code review and questions through Slack or similar. - 3 to 6 months minimum, evaluated quarterly. - IP assignment for any teaching materials, documentation, and reference code you produce specifically for us. Required in your application — answer these five questions specifically. Generic cover letters and AI-generated proposals will not be reviewed: 1.- What's the most substantial production project you've built using Claude Code, the Anthropic API, or MCP servers? Be specific about what you built, what the technical challenges were, and what you learned. 2.- Share a link to teaching content you've created — a video, blog post, GitHub repo with educational README, course material, anything that demonstrates your ability to explain technical concepts. If you've never produced public teaching content, briefly describe a private context where you transferred technical skills to others. 3.- Pretend I'm a 55-year-old commercial real estate broker who runs a 30-person operation, has never written code, but is sharp and wants to learn enough to supervise an AI build intelligently. In 100 words or fewer, explain what an MCP server is and why I should care. 4.- Look at the seven representative projects listed above. Pick one and tell me how you'd build it. Which parts belong in Claude Code, which belong in N8N or pure API calls, and where would you start? You don't need to design the whole thing — show me how you reason. 5.- What's your honest opinion on the current state of Claude Code? What does it do well, where does it fall short, and what's one thing about it that most people get wrong?