Michael K. Onyekwere

What I do

I write Compliance Engineering, a weekly newsletter for engineers and founders who are building AI systems and trying to keep them lawful. Practical, technical, written for people who actually have to ship things.

Through Janus Compliance, I run a small consulting practice. I do paid scoping reviews, DPO-as-a-Service retainers, and AI Act implementation work. The newsletter is free. The consulting is how I keep doing the work.

Why this work

Most developers can build AI but don't understand the law. Most lawyers understand the law but can't read a system architecture diagram. The gap between those two is where most AI compliance failures happen.

I sit in that gap. CIPP/E certified, qualified in law, and writing the system prompts myself. That combination is rarer than it should be, and it's the angle this newsletter is written from.

Background

• Royal Bank of Scotland: financial services compliance
• Fidelity: regulatory operations
• TMF Group: corporate governance and compliance services
• UnitedHealth: health data protection
• CIPP/E: IAPP's privacy certification for European data protection law
• LLB / LLM: common law qualified lawyer

What I've built

AI Shield, a deepfake detection engine using forensic analysis and Claude vision. Three-stage pipeline: signal-level forensics, AI reasoning, arbitration. Built from scratch.

AgentScore, a trust scoring platform for AI agents and MCP dependencies. Full-stack application with policy gates and repo-level memory.

Both ship with proper DPIAs, vendor due diligence, and the documentation a regulator would actually accept. That's the work. Sample deliverables here.

The position

I'm trying to be the person you call when you're building AI and the legal team is asking questions you can't answer, or when the engineering team is asking questions your lawyer can't answer. Not many people do both. That's the gap I work in.