Senior AI Automation Developer — n8n + Retell AI + Next.js (Restoration Industry)

We're K Squared AI, an AI agency building a full AI automation stack for a restoration contractor (water/fire/mold damage). The project has three components that need to work together as an integrated system. We're looking for a senior developer who has built production automation pipelines before — not someone learning on the job. This is a scoped project with potential for ongoing work if the engagement goes well. What We're Building Sprint 1 — CRM-Triggered Email Automation * Albi CRM (REST API v5) webhook triggers on field note / project update events * n8n self-hosted orchestration pipeline: webhook intake → Albi API data enrichment → Claude API for AI email drafting → human-in-the-loop PM approval step → dynamic email send via Gmail OAuth2 (per PM) or Postmark with domain authentication * Emails must send from individual sender identities (name + email), not a system address * Structured XML prompting for Claude with explicit output schema (JSON subject + body) * Audit log of all sent emails written back to Supabase and to Albi activity feed via API Sprint 2 — AI Voice Receptionist (Retell AI + Deepgram Nova-3) * Rebuild existing inbound voice agent on Retell AI using Deepgram Nova-3 for STT * Keyterm prompting configured for service area proper nouns (street names, surnames) * Mandatory double-confirmation loop for name and address capture baked into agent prompt * Structured JSON extraction schema: caller name, property address (street/city/state/zip), loss type enum, confirmed_by_caller boolean * Clear AI disclosure at call open, warm transfer to human agent on request or fallback * Call logs written to Supabase in real time * Integration with existing phone system (TBD) Sprint 3 — Visibility Dashboard * Next.js 15 (App Router) + Supabase + shadcn/ui * Three live panels: email automation log, call log, active job status from CRM API * Supabase Realtime subscriptions for live call monitoring * Supabase Auth, single-user to start with role expansion path * Deploy to Vercel, custom subdomain Hard Requirements — Do Not Apply Without These * Production n8n experience (self-hosted preferred), not just Make or Zapier * Retell AI or Vapi voice agent builds with warm transfer in production * Deepgram STT configuration experience (keyterm prompting, Nova-3) * Next.js 15 App Router + Supabase (RLS, Realtime, Auth) * REST API integration experience — we are connecting to an underdocumented CRM API and you need to be comfortable working from partial documentation * Claude or OpenAI API prompt engineering for structured JSON output * Must be comfortable with ambiguity — some API schemas are pending confirmation from the client's vendor Nice to Have * Prior restoration, field service, or trades industry experience * Gmail OAuth2 multi-user setup or Microsoft Graph API for email sending * Railway or similar self-hosted deployment experience * Postmark or SendGrid domain authentication setup Engagement Details * Type: Fixed-price per sprint or hourly (open to discussion) * Timeline: Sprint 1 target — 2-3 weeks post scoping; Sprints 2 and 3 to follow * Communication: Async-first with weekly check-ins, English fluency required * Budget: $2,500-$4,500 per sprint depending on final scope To Apply Please include: 1. One example of a production n8n workflow you've built involving an external REST API and an AI API call — describe the architecture briefly 2. One example of a voice agent you've deployed in production (Retell, Vapi, or Bland) — what platform, what use case, any accuracy challenges you solved 3. Your hourly rate or fixed-price estimate per sprint based on the scope above Applications without specific examples will not be reviewed.