Lawyer who builds. Founder of AutoJus.
Building the architecture that keeps lawyers in command of legal AI.
I'm a lawyer who spent three decades in practice, then started building the tools I wished existed. My focus now is AutoJus — an agentic legal engineering platform I founded, built on a simple conviction: AI should extend the lawyer's judgment, never replace it. That conviction is architectural, not rhetorical. In AutoJus, the AI proposes; the lawyer disposes — by construction. The agent can draft, research and verify, but the decision to act stays where it legally belongs: with the professional. I build for the lawyer who wants to direct the machine, not defer to it.
Agentic legal engineering platform built on Bring Your Own Key: the firm connects its own accounts across 8 model providers, owns its data path, and pays providers directly with zero markup. Human oversight is enforced structurally — the agent can propose an item to the lawyer's agenda but cannot mark it complete, because that capability was never handed to the model. The AI proposes; the lawyer disposes, by construction. Compliance is a design principle aligned with the EU AI Act. Surfaces: Web, Desktop, Chrome, MCP.
A lawyer commits to diligence, not to a guaranteed outcome — and any AI that behaves as an obligation of results makes a promise no professional can make. The machine does the heavy work of the means; the judgment stays human, because legally it can belong to no one else.
A guardrail written into an instruction is a request the model can ignore. A guardrail written into the architecture is a limit it cannot cross. The difference is a sign that says "do not enter" versus a door that is locked.
A wrong fact you check. An action the machine claims to have taken, you trust — until the deadline passes. I build for the second risk, not just the first.