Agent Systems
Architecture, orchestration, evaluation, memory, observability, governance, and production failure modes.
AI systems research · architecture · technical advisory
Bring me your difficult technical question.
I help founders, investors, and technology leaders evaluate complex AI systems, autonomous agents, agent security, model strategy, privacy, and emerging software architectures. My work is analysis-first: after a short intake, I investigate the problem independently and return with a written, evidence-backed recommendation.
Experience
25+ years in emerging technology and production systems
Enterprise
5 years leading AI/ML & Automation technology business development at Morgan Stanley
Now
Founder & Chief Architect, Pyramidal
Areas of expertise
I work where capable models meet the systems around them: orchestration, evaluation, authority, privacy, interfaces, and technical risk.
Model Evaluation & Routing
Capability boundaries, model selection, structured output, cost-performance trade-offs, and routing strategy.
AI Technical Diligence
Architecture, scalability, defensibility, technical claims, infrastructure economics, and implementation risk.
Privacy & Cryptographic Systems
Zero knowledge, agent authorization, confidential computation, private transactions, and trust minimization.
Emerging AI Interfaces
Generative UI, voice-first systems, intent-driven software, and interfaces that adapt to the question.
AI Agent Security & Containment
Security architecture for agentic systems: least privilege, sandboxing, tool permissions, identity and authorization, secret handling, prompt injection, network egress, auditability, approval boundaries, third-party MCP/tool risk, and blast-radius containment.
AI-Assisted Software Engineering
Architecture, implementation, testing, refactoring, and repository-scale engineering, with AI pair-programming as a daily practice. More in About.
Independent Research
Focused technical investigation for questions that need more than a quick call or generic market scan.
Ways to work together
Fixed-scope engagements designed around decisions, not billable hours.
Transparent pricing
Prices are published so you can tell whether an engagement makes economic sense before we speak. Fixed-scope work is priced exactly; broader engagements show a starting price. Final scope is confirmed before payment.
Second Opinion or Research Brief
An independent review of an architecture, proposal, vendor recommendation, or major AI design decision — or bring one difficult technical question and I'll investigate the landscape, evidence, alternatives, and trade-offs. Includes a short intake, offline analysis, a concise written assessment, and a debrief.
Deep Technical Review
A deeper assessment of architecture, agent design, model strategy, evaluation, privacy, security, operational risk, and implementation trade-offs. Written report plus findings review.
From $5,000
Book fit call
AI Agent Security Review
A focused security-architecture review for teams deploying agents with access to code, infrastructure, credentials, internal data, browsers, APIs, or financial authority. Covers privilege boundaries, sandboxing, tool and MCP exposure, secret handling, prompt injection, network egress, human approval gates, auditability, and blast-radius containment.
From $5,000
Book fit call
AI Technical Due Diligence
Independent technical diligence for investors, acquirers, boards, and corporate development teams. Covers what actually exists, architecture, differentiation, economics, technical claims, risks, and questions for management.
From $7,500
Book fit call
Advisory Call
A focused 60-minute conversation for situations that genuinely benefit from live discussion rather than a written analytical engagement.
$750
Book Advisory Call
Research. Review. Recommendation.
The core product is not my time on a call. It is a well-researched answer to a difficult question.
The core product is not my time on a call. It is a well-researched answer to a difficult question.
Already know what you need?
Skip the fit call. Send me the problem, the decision you need to make, any relevant materials, and your preferred timing. If the scope is a fit, I’ll confirm it by email and send the payment link.
Selected work
My advisory work comes from active technical practice, not distance from the work.
Founder building products and open standards around AI systems, privacy, and autonomous agents.
Boundary-driven model evaluation and routing based on where models stop working, not just where they rank.
Verified and qualified professional contact designed for an era of near-zero-cost AI-generated outreach.
A voice-first interface whose workspace assembles around the user's intent and data.
Privacy-preserving settlement, compliance, delegated authority, and cross-organizational agent federation.
Measurements, proofs, inspectable architecture, and written analysis over unsupported claims.
Writing & research
After Intelligence explores what matters when capable AI becomes abundant: discernment, judgment, agency, and the systems that turn capability into consequence.
Editorial & research
After Intelligence
A home for the questions that begin after the model works: what deserves attention, who can act, on whose authority, and what people should be able to trust.
Explore After Intelligence →First major series · Eight essays
Privacy, Power & the Machine Economy
How privacy, delegated authority, and compliance can support autonomous commerce without turning shared infrastructure into an instrument of control.
About
I am the founder and chief architect of Pyramidal and work across AI systems, autonomous agents, agent security, model evaluation, privacy, cryptographic infrastructure, and adaptive software interfaces.
Before Pyramidal, I spent five years at Morgan Stanley leading AI/ML and automation technology business development, evaluating emerging technologies, advising on product recommendations, and helping connect promising technologies with enterprise adoption and investment opportunities.
My work in machine learning and data mining stretches back to the 1990s, and for nearly four years I have practiced AI-assisted software engineering as part of my day-to-day development process. Codex and Claude Code became part of that workflow after their 2025 launches. The work spans architecture, implementation, testing, refactoring, and repository-scale development.
My focus is increasingly on dependable autonomy: the infrastructure required when AI moves from answering questions to making decisions and taking actions in the world.
I take on a small number of independent technical advisory engagements alongside my work at Pyramidal.
Have a difficult technical question?
If it involves AI architecture, autonomous systems, agent security, technical diligence, model strategy, privacy, or emerging AI infrastructure, there are two easy ways to start.
Advisory engagements are provided and billed by Pyramidal.