About
I build the layer between raw information and a decision.
I'm an intelligence systems engineer. My work is designing, deploying, and operating AI-driven systems that ingest information, organize it, reason over it, and turn scattered data into intelligence someone can act on.
The work
My offer is intelligence infrastructure: research engines, document ingestion systems, and applied AI products that turn messy information, scattered documents, web data, and underused knowledge into something structured, searchable, and actionable. The form changes by customer. The core value does not.
I treat AI as infrastructure, not a chatbot. The interesting problem is rarely the model. It is the system around it: how content gets in, how it is structured, how it is retrieved, and how an answer arrives with its sources attached.
The arc
Every project I have shipped is a stage of the same system, not a separate product. The through line is intelligence amplification: how do you build a machine that finds what matters before everyone else does?
Triager+ proved I could take a model the whole way to a usable product. ShapeSound was a study in language design, parsers, and symbolic systems. Manifest is the graduation piece: a document intelligence engine and the foundation everything else now runs on. Aletheia adds the reasoning layer, and verticals like Playbook apply the intelligence to a specific market.
Build the foundation once. Every vertical inherits ingestion, embeddings, search, and sourcing, and adds only what its domain needs.
Where this is going
Manifest, Aletheia, and the verticals are being built as portfolio-grade systems, and the progression is documented here in public. The aim is a foundation strong enough that every vertical built on it is worth standing up a business around.
Infrastructure over demos
A demo impresses once. Infrastructure compounds. I build the part that keeps paying off.
Sourced, not guessed
Every answer carries its citations. If it cannot be traced to a source, it is not intelligence.
One foundation, many verticals
Solve the hard part once. Let each domain inherit it and add only what is unique.
Ship the edges
A system is clearest at its boundaries. Scope tight, finish it, then expand.
Want the longer version?
Read the thesis in full, or see the architecture proven in code.