I build intelligent systems that work.
Agent orchestration, conversational analytics, machine learning platforms, semantics, data, and the mathematics underneath them.
From intent to execution.
Architecture before hype.
I’m interested in the parts of AI systems that become important after the demo works: semantics, boundaries, routing, reliability, observability, security, and cost.
Blu Agent Orchestrator
Capability discovery, policy filtering, agent routing, A2A delegation, MCP-backed execution, and routing evaluation.
Semantics before SQL.
The user sees conversation. The platform needs semantics. Business meaning should be governed, executable, and testable.
Understand the idea. Then write the code.
Mathematical Foundations for Machine Learning
Linear algebra, probability, optimization, regression, PCA, SVD, embeddings, and neural networks taught in the order I think they should be learned: intuition → geometry → mathematics → code → application.
Explore the course →[ 1 3 ]
Notes from the workbench.
Ideas I want to understand well enough to explain without hiding behind jargon.
Agent orchestration is not an MCP routing problem
Why registry, policy, routing, and tool access belong to different layers.
Why semantic layers matter when users stop writing SQL
Removing syntax from the interface makes governed meaning more important.
What eigenvectors are actually telling you
A geometric explanation before the algebra starts.
Your query isn’t just code. It’s a bill waiting to happen.
Why efficient data systems are an engineering and economic concern.
Understand. Build. Teach.
I’ve spent more than two decades building software and more than a decade working with machine learning. This site is where I document systems, experiments, lessons, and courses that are worth keeping.