Both remote. Both long-term. 2-week paid trial. Indefinite contract after if it's a fit. Equity conversation at month 6 for strong performers. Role 1: Implementation Lead (DWY + DFY Delivery) You own delivery end to end. We close, you ship. What you'd actually do: 3–5 Zoom calls a week with new and active clients (CEOs, ops leads, media buyers)Build custom Claude Code workspaces in their Drive: CLAUDE.md, slash commands, MCP servers, ClickUp/Slack/Gmail integrationsTrain each user on 3 real workflows for their roleWrite a recap doc and Loom for every installFirst responder for 30-day post-install Slack support You'll thrive here if: Native or fluent English — you'll be on Zoom with non-technical CEOs dailyUS / Canada / UK / AU / South Africa with overlap into US EasternYou've used Claude Code in anger: you can write a CLAUDE.md, build a custom command, and connect an MCP from scratchMac terminal comfort: brew, npm, .env, basic gitYou can walk a 55-year-old CEO through cd ~/workspace && claude without making them feel dumbBonus: prior CS / Implementation / Solutions Engineer role, or hands-on delivery work on Make, n8n, Zapier, GHL Apply → https://docs.google.com/forms/d/e/1FAIpQLSdSPGwN7zPo-WGbtGlOaWBmOysDqeXHkMk9zPh_kHHs4OY7bQ/viewform Role 2: Senior Full-Stack + AI Engineer Pure engineering. No client-facing. Replacing a senior contractor who's rolling off. Stack: Python, FastAPI, Postgres, React, Next.js, TypeScript, Tailwind, Claude API (SDK + tool use + MCP), OpenAI, Nvidia Nemotron, VAPI, Twilio, Azure Speech, ElevenLabs, ffmpeg, Vercel, Docker, Hetzner, Azure. You'll thrive here if: You've shipped production FastAPI services AND React frontends — bothYou've integrated VAPI, Twilio, Vocode, or Retell in a real productYou can read a slow query plan, rewrite the schema, and ship the migration the same dayYou've used ffmpeg from the CLI in a real product (not a tutorial)You don't need a PM — you read the ticket, ask one clarifying question, and ship

A UK investment firm lost $1.2 million in Q1 because their AI hallucinated a merger announcement that never existed. Hospitals suspended AI trials because models were flagging patients for diseases they didn't have. Even the "best" models still hallucinate 17-72% of the time depending on the task. So here's what I've been doing: - Flags uncertain responses for human review - Uses RAG (Retrieval-Augmented Generation) to ground answers in real data - Reduces hallucination by ~75% when web search is enabled - The reality? AI is powerful but NOT reliable on its own yet. The money is in building RELIABLE systems, not just fast ones. How are YOU handling AI accuracy in your client projects? What verification steps do you use?