Production AI Agents

I do not build or train foundation models. I build the ecosystem, tools, integrations, and orchestration around them so they can take real action.

8 autonomous agents deployed in production across sales, CS, support, and marketing.
Self-hosted infrastructure running entirely on Hermes (open-source by Nous Research).

Engineering Output

291 PRs merged autonomously in the first 3 months.

Issue Resolution

171 issues closed by the engineering agent.

Cost Optimization

Ops agent cut 15% of AWS costs.

Workflow Shift

Planning moved from Notion/Jira to GitHub (agents work where the code is).

Under the hood

Strategic Governance

Weekly OKR review cross-checks the board against the quarter's key results, flagging epics with no linked KR.

Delivery Pipeline

Backlog runs on GitHub Projects — Sprint and Epic boards — plan, issue, PR, same system as the code.

Self-Healing

A cognitive audit reviews the agents' own memory and skill usage, opening issues against itself to refactor.

State as Code

Runtime config and memory sync to GitHub via automated PR twice a week.

Security Watchdogs

Two independent audits scan for vulnerabilities every morning; only a nonzero finding pages the team on Telegram.

If you're running AI agents in production — where a mistake is expensive and delegation needs to be real, not nominal — write to me.