AI Systems Architect · Constraint-Native Architecture· Regulated-Domain AI

I build agentic AI systems for high-stakes, regulated industries production deployments where safety, compliance, and auditability are structurally baked into the architecture, not retrofitted after the fact.

I coined the term Constraint-Native Architecture: a methodology where regulatory, privacy, and safety constraints are the generative logic behind an agent's design not a governance layer added afterwards. The system can't violate constraints because it was never designed to permit that path.

I hold credentials from Johns Hopkins (Agentic AI), Oxford Saïd Business School (AI Ethics, Regulations & Compliance), and oxethica (Certified AI Auditor · Certified AI Ethicist · Certified AI Professional) one of the few practitioners combining hands-on production agentic systems with formal audit and governance training.

• Production agentic AI for pharma strategy, financial compliance, and regulated-domain workflows
• Constraint-Native Architecture patterns: PII non-acquisition by design, RegOntology-first agents, tamper-evident validation graphs
• Knowledge graph-driven agent topologies (Neo4j + hybrid retrieval)
• Reusable AI developer toolkit RAG pipelines, validation frameworks, multi-agent orchestration
• Interactive data storytelling: D3.js knowledge graph visualizations and automation control centers

• Collaboration on regulated-domain AI (pharma, fintech, health tech, legal)
• Speaking and advisory conversations on agentic AI governance and safety architecture
• Practitioner co-authorship on agentic systems design patterns
• Consulting engagements in AI governance, audit readiness, and compliance-native system design

"Architecture should make unsafe behavior structurally impossible not just discouraged."

I approach agentic systems by decomposing the problem space before writing a line of architecture. The systems I design are defined as much by what they cannot do as by what they can and every autonomous decision is a verifiable, auditable event, not a log entry.

I work from negative space: I define what the system must never do, must never store, must never infer without permission and the resulting design emerges from those constraints rather than fighting them. This produces more compliant, more robust, and more trustworthy systems than any post-hoc governance layer can achieve.