Software engineer building forensic systems, governed AI, and product architecture for high-trust, high-stakes environments.
My work sits at the intersection of engineering and trust. I focus on systems where correctness is not optional, where outputs need to be defensible, and where the gap between “it works” and “it holds up” determines whether the product matters.
I am a founder and systems-focused engineer. I build GhostLogic, a forensic evidence platform, and Maelstrom Runtime, a governed AI orchestration system. My product work spans early-stage architecture, technical prototyping, and full-stack delivery.
I care about structure, failure modes, boundaries, and building software that behaves reliably under real conditions. I work best in environments that are technically dense, high-trust, and commercially important.
Systems built for trust, traceability, and real-world execution.
Runtime AI Governance
Runtime governance for tool-using AI agents.
Same model. Same task. Different threat level. Dangerous tools disappear from the model's visible action surface under elevated risk.
Open source on GitHub. Live demo available. Paid pilots now open.
Forensic Infrastructure
A forensic evidence platform that captures, seals, and verifies digital artifacts for legal and insurance use cases. Multi-component system spanning telemetry collection, evidence capsule storage, forensic analysis, and claims settlement.
Governed AI
A governed AI orchestration runtime that enforces behavioral constraints, maintains audit trails, and ensures model outputs are reproducible and accountable. Built for environments where AI decisions carry real consequences.
Architecture & Delivery
Product architecture, technical prototyping, and full-stack delivery for early-stage companies and complex technical problems. Turning ambiguous requirements into working software.
Five areas of focus, unified by a commitment to systems that work under pressure.
Evidence infrastructure built for integrity, traceability, and legal defensibility. Systems that capture, seal, and verify digital evidence under adversarial conditions.
AI orchestration with built-in governance, audit trails, and behavioral constraints. Making model outputs accountable and reproducible.
Turning ambiguous requirements into coherent technical systems. Defining boundaries, data flows, and interfaces that survive contact with real users.
End-to-end execution from architecture through deployment. Building the thing, not just designing it. Shipping code that works in production.
Helping teams and founders make sound technical decisions early. Stack selection, build-vs-buy analysis, infrastructure planning, and roadmap clarity.
Five principles that shape everything I build.
If you cannot explain what a system does and why, it is not ready to ship.
Good systems are shaped by their boundaries, not just their features.
Every decision, every state change, every output should be explainable after the fact.
What happens under load, under failure, under adversarial conditions is the real design.
Architecture without execution is just a diagram. The work is not done until it runs.
I tend to work with people navigating complex, high-stakes technical problems.
Founders who need a technical co-builder, not just a developer.
Operators running systems where failure has legal, financial, or safety consequences.
Technical teams working on problems too complex for off-the-shelf solutions.
Companies building products that require evidence integrity, AI governance, or regulatory compliance.
Early-stage startups that need architecture decisions made well, the first time.
Legal and insurance professionals who need technology that produces defensible, auditable outputs.
Founder Post
Most AI safety still depends on the model refusing to do something dangerous. I think that's the wrong layer.
Maelstrom Gate governs the model's available action surface itself. Under elevated risk, dangerous tools vanish from context before the model can select them. No refusal theater. No jailbreak argument. Just control upstream of action.
The enforcement primitive is open source. The broader runtime and pilot work are live.
I am always interested in hearing about hard problems. If what you are building requires real engineering rigor, let's talk.