About

Laughing Man

An AI consulting and engineering practice. Every engagement is led by Mike Lavelle.

Mike Lavelle, principal

Mike is a senior platform engineer who works AI-native: agents do most of the typing, and he builds the systems that decide when their work is done. He uses AI for nearly everything he does, from code to reports, research, and requirements. Mike's own Claude Code usage runs at more than 30 times Anthropic's published per-developer average.

He has done the jobs agents are now being asked to help with: customer support, QA, requirements gathering and specs, process writing, engineering, DevOps, and identity and access management (IAM, workload identity federation). He runs infrastructure on Azure and Google Cloud as code, and has worked with PostgreSQL, Supabase, MongoDB, and most of Azure's services. He was also the technical liaison for a national predictive-modeling initiative, working with a national lab, a federal research agency, and more than ten universities, and explaining the work to members of Congress.

On the AI side

  • Local and private models. LLMs that run entirely on hardware you control, so sensitive data stays on it.
  • Fine-tuning, retrieval (RAG), and MCP servers that connect models to the systems a business already runs.
  • Agent evals, so a change to a prompt, a skill, or a model shows up as a measured difference, not a hunch.
  • Agents and skills that improve with use. Once one works at a basic level, every failure becomes a test or a rule, so the same mistake is caught the next time.

That range is the point. Mike has been the support rep who hears what's broken, the engineer who ships the fix, the QA lead who has to sign off, and the admin who owns the access. We fit AI into a team knowing what each of those people needs.