AI Architect & Automation Consultant

Thanks for all the interest. I will be reviewing proposals over the next week or so. Please stay tuned! Travis Hedrick Founder + Serial Problem Solver @ iComm Studios == ABOUT US == iComm Studios is a small but mighty digital marketing and web design agency in North Carolina. Like most folks in our industry, we are actively planning how and where AI fits into our current and future business models. Spoiler Alert: I have no formal training and limited knowledge. I'm quite surprised we made it this far. Now we need to tighten up our stack and build the foundation of our upcoming business automation sector with the help of a skilled AI architect and consulting partner. == CURRENT STACK == We are not starting from scratch. We run a functioning, self-hosted Hermes stack that I built blindly, and it already saves an exponential amount of time and has multiplied our quality level and customer value. It has live integrations across client data sources (Google Ads, GA4, GSC, Monday.com, Fireflies, Google Drive), cron-based automation, a persistent memory layer (Mnemosyne), a custom skill library, and multi-tool orchestration across agency tasks. What we don't have yet is a technical partner who has seen enough of these stacks to tell us where our architecture decisions will cause problems before they do. == LONG-TERM PARTNERSHIP == We want a long-term partner, starting with a working audit of our current setup, then ongoing consultation on specific questions as they come up. If things go well, phase two will be rolling out our agentic digital marketing options as well as an AI sector that focuses on business automation. The topics we expect to cover are grouped below. == TECHNICAL == Token and prompt caching Whether we are configured to capture API-level savings and where we are leaving money on the table. Model routing and tier logic A second opinion on our 4-tier model map (Haiku through Opus) and where the lines are drawn. Memory layer design Consolidation strategy, scope hygiene, and retrieval accuracy in Mnemosyne. Context window management Cron jobs, background agents, and per-turn injection all push context size; we want proven patterns to reduce overhead. Security baseline Credential handling, hook enforcement, and secret exposure risk in a local-first production environment. Alerting reliability Architectural options for surfacing delivery failures in watchdog and cron jobs. Integration architecture Connecting Slack, Google Workspace, Monday.com, Google Ads, GSC/GA4, and third-party APIs reliably. Log analysis and anomaly detection Automated triage, false positive reduction, separating actionable signal from noise. Upstream update management Carrying local patches against a fast-moving upstream without drift. Agent orchestration Claude Code CLI vs. delegate_task routing, multi-agent Kanban patterns, stall detection. == BUSINESS PROCESS AUTOMATION == Data analysis and manipulation Pulling from GA4, GSC, Google Ads, and client reporting systems, then normalizing and summarizing for agency use. Quality assurance and review Automated content review pipelines, AI detection thresholds, approval workflows, preview delivery and nudge logic. Project management workflows Monday.com board updates, task creation from triggers, status reporting, client milestone tracking. Client onboarding and briefing Intake automation, context card generation, document filing, Drive organization. Content production pipelines Blog, social, and campaign content from brief to delivery, including humanization and AI detection QA gates. Client deliverable management Branded document production, Drive filing, version control, client-facing output QA. Reporting automation Scheduled performance summaries across client accounts, delivered to the right person at the right cadence. == IDEAL FIT == Hands-on experience with the Hermes Agent framework or an equivalent Anthropic-based agentic stack, a solid grasp of prompt caching at the API level, and someone who treats security as a first-class concern from the start. ENGAGEMENT START: ASAP SCOPE: Initial working audit session; written findings preferred.