Maryam Miradi - DeveloPassion

# Maryam Miradi Maryam Miradi is an AI scientist and educator whose work centres on **AI agents in production** rather than in demos. She holds a PhD in AI and describes 20+ years in the field. ## Why she's in my vault Her four-loop framing of [[Loop Engineering]] is the clearest version of that idea I've come across, and it contributes something the other taxonomies mostly leave out: 1. **Task execution** — plan, act, observe, repeat until done 2. **Verification** — output checked against tests, rules or evals, not the agent's self-assessment 3. **Orchestration** — triggered from outside (Slack, GitHub, email, queues, schedules), running end to end 4. **Continuous improvement** — analyse traces across *many* runs, find repeated tool errors and weak prompts, then fix the prompts, tools, skills, evals or workflow Loop 4 is the contribution. Everyone builds the first three; almost nobody builds the one that operates on the population of runs rather than a single run. Her argument for it is sharp: without it, one repeated failure doesn't stay a small bug, it scales across thousands of executions. She also places human approval *inside* any of the four loops rather than treating oversight as a separate stage, which is the better framing because it makes approval a property of a step instead of a phase that can be skipped. ## Her focus Production AI agents, taught through real-world projects across healthcare, legal, cybersecurity, finance and supply chain. Her material covers [[Claude Code]], LangGraph and [[Model Context Protocol (MCP)|MCP]]. She publishes a newsletter, *AI Horizon Watchers*, and free guides including a 56-page "50 AI Agents Real-World Best Practices". The through-line worth noting: her emphasis is consistently on what breaks when agents run at scale and unattended, which is the gap between the demos most people write about and the systems that survive contact with production. ## References - [maryammiradi.com](https://maryammiradi.com) - [LinkedIn](https://www.linkedin.com/in/maryammiradi/) - [YouTube](https://www.youtube.com/@maryammiradi) - [X](https://x.com/MaryamMiradi) - [GitHub](https://github.com/maryammiradi) - "The Art of Loop Engineering" (2026-07) — https://www.linkedin.com/posts/maryammiradi_the-art-of-loop-engineering-how-to-share-7487945874938077185-1gB_ ## Related - [[Loop Engineering]] - [[AI Agents]] - [[Agentic Engineering]] - [[AI Verifiability]] - [[Model Context Protocol (MCP)]] - [[Claude Code]] ## Quotes  ## Books