← Expertise
AI agents & multi-model automation
I design and run OpenClaw — a multi-model operations assistant that routes work across Claude, GPT and Gemini, reachable from Discord and WhatsApp, with approval gates so it never ships junk autonomously.
What I build
01Multi-model routing — the right model per task
02Agent / skill architecture and tool access
03Approval gates for autonomous work
04Chat-surface interfaces (Discord, WhatsApp)
Selected work
Shipped, in production.
Common questions
Can you build an agent that uses my own tools and data?
Yes — that's the point. OpenClaw connects models to real tools and approval flows. I build the routing, the skills, and the guardrails so it's useful without being reckless.
Which model do you use?
Whichever fits the task — I default to the latest Claude models for reasoning and code, and route to others where they're stronger or cheaper.