- Built the company's single source of truth from nothing across a portfolio of 112 active clients — designed a normalized, governed ~30-table Airtable base as the canonical system of record.
- Architected P3 Command Center (→ read the case study): Airtable as source of record, an operational cache & service layer, and the marketing CRM as the execution layer — unified via custom REST/JSON APIs.
- Built production AI workflows — Claude & Gemini agents, embeddings/RAG retrieval, MCP tool access — all with human-in-the-loop approval gates; and an AI-discoverability audit tool that scores pages for LLM-driven discovery.
- Replaced scattered, ad-hoc LLM chats with a centralized content engine: client profiles auto-researched (web scraping + scheduled SEO-data pulls + AI), then routed and tuned per channel — a Clay-style enrichment engine built with no-code platforms and APIs.
- Single technical point of contact across the portfolio — WordPress, client CRMs, Google Business Profile, Moz Local, Google Ads, and Meta asset ownership — translating non-technical owners' needs into shipped solutions and training them to adopt the new workflows.
- Reliability & security hardening: stale-while-revalidate cache, retries, durable webhook retry queues, schema-drift detection, and an Airtable write lockdown (write logic extracted to a service layer). 0 type errors, 920/923 tests passing.
- Conservatively estimated 50-80 hours/month reclaimed across a six-person core team through centralized status, cleaner approvals, better reporting context, and more reliable integrations. (Internal estimate, not an employer-tracked KPI.)
AirtableMake.comGoHighLevelClaude / GeminiRAG / embeddingsMCPREST / JSON APIsWordPress