AI Automation Engineer: Multi-Cluster Newsletter Engine & Historical iCloud Archive Miner

AI Automation Engineer: Multi-Cluster Newsletter Engine & Historical iCloud Archive Miner (Gmail/iCloud ➔ Airtable/Notion) *Must be fluent in English, friendly, efficient, trustworthy and communicative. Project Overview I am building a comprehensive content repository to eventually power an app, a podcast, and my own newsletter. This content is sourced from high-quality email newsletters across several distinct topic clusters (AI, Finance, Health, etc.). I am looking for a talented automation and AI prompt engineer to build a multi-phase system. This system must process both incoming daily newsletters and two years of historical archives (stored in iCloud), route them through specialized AI models tailored to each topic, and funnel approved insights into a pre-categorized master database (Airtable or Notion). Detailed Scope of Work Phase 1: Multi-Cluster Daily Processing (Incoming) * Trigger & Segmentation: Monitor incoming newsletters in Gmail. Based on the sender or Gmail labels, the system must split the emails into parallel topic workflows (AI, Finance, Health). * Specialized AI Prompting: Pass the batched 24-hour text to an LLM (OpenAI, Claude, or Gemini API). Each cluster must use a unique, hyper-targeted prompt rule (e.g., the AI prompt hunts for code repos and tech tutorials; the Finance prompt hunts for market metrics and VC funding; the Health prompt looks for clinical trials and biohacking protocols). * Crucial Link Extraction: The AI must be strictly instructed to retain original HTML/markdown hyperlinks for any featured tools or websites mentioned so they remain fully clickable in the final output. * Delivery: I want to choose between receiving staggered, topic-specific email digests, or routing the summaries directly into dedicated topic dashboards within my database. Phase 2: Categorized Historical Mining (The "iCloud Backlog Sweep") * Target: I have two years' worth of newsletters organized into folders inside iCloud. * Action: Build a one-time batch workflow to mine these historical archives. The system must use the same specialized AI topic criteria to extract data and map it retrospectively into the correct topic clusters. * Cost Optimization: Because this involves thousands of old emails, I need a developer who can process this cost-effectively (e.g., using a custom Python script or batched API calls rather than triggering thousands of individual multi-step Zapier tasks). Phase 3: Pre-Sorted Database Architecture (Airtable or Notion) * I do not want a traditional Google Spreadsheet. I want to use Airtable or Notion to act as my relational content hub. * Topic Routing: The database must be structured with dedicated views or connected tables for each cluster (AI, Finance, Health) so content is automatically pre-tagged and categorized upon arrival. * "Human-in-the-Loop" Approval Logic: Build a quick verification layer (using Relay.app checkpoints or a "Review Status" dashboard) where I can easily review the AI's extracted tips and check a box to instantly push them into my Master Library. Technical Stack Requirements You should be highly proficient in: * Automation Engines: Relay.app, Zapier, or Make.com * Databases: Airtable or Notion (Relational database design, multi-view dashboards) * AI/LLMs & Prompt Engineering: Advanced prompting using OpenAI, Anthropic/Claude, or Google Gemini APIs * Protocols & Scripting: IMAP configuration (for the iCloud email connection) and Python/NodeJS for low-cost batch processing. How to Apply (Screening Questions -ANSWER ONE or MORE) Please provide a brief, customized response addressing the following: 1. The iCloud Integration: Since iCloud doesn't have a native 1-click connector in Zapier, what is your proposed method for mining the last two years of newsletters sitting in my iCloud folders? 2. The Multi-Cluster Challenge: How will you structure the automation workflow so we can easily add new topic categories in the future without rebuilding the system from scratch? 3. The Link Retention Problem: How will you ensure the AI successfully extracts and retains the direct website hyperlinks from the newsletters so they stay clickable for me in the final database? 4. The Backlog Sweep: How do you propose we run the historical archive processing to keep API/Automation platform costs as low as possible? 5. Database Suggestion: For an app, podcast, and newsletter engine split into distinct topic areas, do you recommend Airtable or Notion for this specific project, and why? Scope of Work: Multi-Topic AI Newsletter Content Engine Objective To build an automated, AI-driven content pipeline that ingests daily email newsletters and extracts specific high-value insights (tools, tips, stories) categorized into distinct topic clusters (AI, Finance, Health). The system will also process a two-year historical archive of newsletters stored in iCloud, funneling all curated content into a centralized relational database (Airtable or Notion) with a "Human-in-the-Loop" approval step. Milestones & Deliverables Milestone 1: Database Architecture Setup (Airtable or Notion) * Deliverable: A fully configured master repository structured for scaling into a future app, podcast, and newsletter. * Requirements: * Create separate tables or dynamic, filtered views for each topic cluster (AI, Finance, Health). * Build a temporary "Review Dashboard" tab where incoming extracted items wait for approval. * Include fields for: Content Source, Date, Topic Tag, Core Summary/Tip, and Original Hyperlinks. * Set up a checkbox or single-select status field (e.g., "Approved", "Rejected") that triggers the transfer of items from review to the Master Library. Milestone 2: Live Ingestion & Multi-Cluster AI Pipeline (Daily) * Deliverable: A live automation workflow (using Relay.app, Zapier, or Make) processing incoming daily newsletters. * Requirements: * Monitor a dedicated Gmail label (e.g., Newsletters) and batch incoming emails over a rolling 24-hour window to create a single daily processing event. * Route batched emails through conditional logic to apply hyper-specific AI prompts depending on the topic cluster (e.g., the AI prompt hunts for software tools; the Finance prompt hunts for market data). * Strict Link Retention: Force the LLM to extract and format direct website hyperlinks as clickable markdown/HTML links inside the database. * Route the final output straight to the database's "Review Dashboard." Milestone 3: Historical Backlog Archive Sweep (iCloud Migration) * Deliverable: A one-time, budget-conscious batch process that extracts data from two years of historical newsletters. * Requirements: * Establish a secure connection to the client's iCloud email archive (via IMAP integration or local script file ingestion). * Process thousands of archived emails through the same multi-cluster AI filtering logic. * Cost Optimization: Implement this via a custom script (e.g., Python) or a highly batched API workflow to avoid excessive multi-step automation task charges. * Dump all historical insights directly into the relational database for review. Milestone 4: Testing, Refinement, & Handover * Deliverable: Operational testing and system handover. * Requirements: * Conduct end-to-end testing with sample live incoming emails and historical archive sets. * Verify that direct website hyperlinks remain fully functional and clickable inside Airtable/Notion. * Provide a brief video walkthrough (Loom) or a 1-page document explaining how to adjust AI prompts, add new topic clusters, and rotate API keys in the future.