CHARLES JONES

Solo builder-operator engineering agentic AI systems on Google Cloud Platform (GCP): multi-agent orchestration, Model Context Protocol (MCP) infrastructure, and two production web platforms totaling 1.35M pages — owned end to end across architecture, build, deploy, CDN/cache behavior, SEO, analytics, security, and uptime. AI agents write most of my code; my job is context, architecture, review, and holding the bar on what ships. Every deploy is manifest-hashed and gate-verified before it goes live — no merge on vibes.

• Custom MCP server, designed and shipped. 26-tool (measured 2026-08-08) Model Context Protocol (MCP) server on Cloud Run, Google Cloud Platform (GCP), serving session state, ledger queries, BigQuery-backed search analytics, and task management to every agent seat — the tooling layer that lets one engineer run a 1.3M-page surface.
• AI-first development at production stakes. Run multiple Claude agents in parallel across dedicated build, deploy, and tooling seats — one implementing, one auditing, one on infrastructure — with myself as the single review gate. Built the governance to make that safe: append-only decision registers, gate verdicts bound to artifact manifest hashes, SHA-verified deploy read-backs, honest-failure reporting (a failed read never renders as a clean line).
• Document/data ingestion at scale. Parsed Florida's public licensing corpus — 688,000+ producer records, 28,000+ agencies — from messy government exports into a deduplicated, structured directory with per-record pages, facets, and schema.
• Licensing-data corpus at national scale. Maintain the full active Florida DBPR contractor-license corpus powering callcharlie.co; currently sourcing nationwide active NIPR insurance licenses and every state's active insurance-license records for national expansion of the directory platforms.
• Funnel and demand instrumentation. Google Search Console BigQuery export analysis (SQL over impression/click/query tables), Google Ads API Basic Access with Keyword Planner via API for first-party demand data, indexation forensics, attribution-safe lead-capture design.
• Production incident ownership. Diagnosed and remediated a compromised service-account credential incident end to end — forensics, key rotation to keyless service-account architecture, IAM lockdown, provider escalation through to a written determination and full credit. Case covered by The Register (July 2026). Closed a Firestore security-rules vulnerability with claim-checked rules and live production verification.