I turn engineering judgment into workflows AI agents can execute.
I'm Rom Iluz. I practice and teach Harness Engineering, Workflow Engineering, and Loop Engineering — the three disciplines that turn AI coding agents from unpredictable one-shots into systems that produce reliable work. I didn't just define the ideas. I built the proof: CC10x, Auto-Pi, Memongo, ClawMongo, and 60+ other open-source repos used by thousands of developers.
The work.
Each project exists because I identified a specific failure mode in AI agents and built the fix. Not tutorials. Not demos. Production tools I use every day.
Each project fixes a specific failure.
I don't build random tools. Every repo exists because I watched AI agents fail in a specific way, understood why, and built the mechanism that prevents it. These are the projects that proved the ideas work — not in theory, but in production, with real users, real stars, and real bugs that became real fixes.
CC10x
The problem: agents declare work complete when it isn't.The fix: one router, nine specialist agents, sixteen skills, four evidence-gated workflows. The agent can't grade its own homework — an independent reviewer with zero context checks every plan and every build. Fail-closed gates, bounded remediation loops.
View on GitHub →
Hybrid-Search-RAG
The problem: pure vector search misses exact matches, and pure keyword search misses meaning.The fix: a hybrid retrieval system that does both and falls back gracefully when one path fails. Built for production RAG pipelines where getting the wrong document is worse than getting no document.
View on GitHub →
WhatsApp AI
The problem: most AI chatbot demos break the moment real users send real messages.The fix: production agents handling actual WhatsApp conversations at scale. Every edge case — typos, voice notes, group chats, media — is a lesson you can't learn from benchmarks.
View on GitHub →
Memongo
The problem: file-based agent memory rots. Agents forget, contradict themselves, and repeat mistakes they already made.The fix: MongoDB-native long-term memory with vector search, episodic memory, and durable state that survives across sessions — so the agent gets smarter, not dumber, over time.
View on GitHub →
ClawMongo
The problem: personal AI assistants lose context across channels and sessions.The fix: a MongoDB-native gateway that answers on WhatsApp, Telegram, Slack, Discord, and 15+ other channels — with one canonical memory backend. Hybrid retrieval, automatic embeddings, real agent memory tools. Fork of OpenClaw, rebuilt for production.
View on GitHub →
Auto-Pi
The problem: if your harness only works on one agent host, it might be a product feature, not a discipline.The fix: the same harness and workflow ideas from CC10x, rebuilt on Pi — a completely different coding agent. Same system, different host. The pattern transfers. 10 packages, 59 skills, 10-step autonomous workflow.
View on GitHub →How I think about agentic engineering.
The model is just capability. A better model doesn't fix unreliable work — it produces unreliable work faster. What makes an agent trustworthy is the system around it.
Harness Engineering controls what the agent sees and does: context, tools, authority, skills, memory, evidence requirements. Curated, not dumped. Enforced structurally, not by asking nicely.
Workflow Engineering is the discipline I practice and teach: taking real engineering workflows — your team's judgment, constraints, domain expertise, review processes — and transforming them into executable contracts that AI agents must follow. Not prompts. Not suggestions. Contracts with stages, evidence requirements, and exit conditions.
Loop Engineering uses evidence to decide what happens next. Not retries. Convergence loops. Claim → evidence → verdict → next state. The system advances, repairs, re-plans, stops, or escalates based on proof — not the agent's confidence.
I built CC10x and Auto-Pi to prove this works on different hosts. I built Memongo and ClawMongo because agents without durable memory repeat their own mistakes. I built Hybrid-Search-RAG because retrieval quality is the difference between an agent that helps and one that hallucinates. Each project is a piece of the same system — and the system is the point.
Want to work together?
I'm available for keynotes, workshops, and advisory engagements. I also build things in the open — follow the work, fork the repos, or reach out directly.