Governing AI. Tracing contributions. Building the next research chapter.

The research I want to pursue

My current research centers on two connected questions: how can AI systems remain accountable to human authority, and how can we trace and govern the intellectual contributions used in AI-assisted work? These are the questions I want to develop through academic collaboration at George Washington University.

I bring a background in computer science, national-security technology, biomedical research, teaching, and company building. My aim is to connect that experience to a focused empirical program: controlled studies, transparent evaluation, and findings that explain when a system deserves trust and when human intervention is needed.

Spyris: from technology transfer to attribution and governance

The first version of what became Spyris grew out of our contract work for the National Security Agency’s Office of Research and Technology Applications (ORTA). The assignment brought together the elements of an end-to-end IP attribution and governance engine: discovering relevant technology, connecting work to its sources, preserving provenance, and supporting review and authorized use.

That origin matters to my research. Technology transfer is a setting in which attribution must lead to a decision: which source is relevant, what evidence supports the relationship, who can evaluate or use the technology, and what permissions are needed? Spyris’s core platform and attribution extension provide a practical starting point for investigating those questions.

The documented collaboration record includes a Spyris Federal–NSA cooperative research and development agreement signed in March and April 2026. It identifies the platform and attribution-engine background and a research scope involving integration, usability, performance, and software assets. This gives the academic agenda an existing body of engineering work and a set of concrete questions to test.

AI IP attribution: evidence that people can evaluate

Through Spyris, I am studying attribution across human edits, model responses, retrieved material, software, and tool use. I want the evidence to explain how a contribution arose and how confident we should be in that explanation.

My proposed experiment uses inventors and domain experts to assess technical relationships, identify supporting evidence, and record reasons and uncertainty. Independent review and held-out source families would help test whether the approach generalizes. A second study would compare attribution alone with attribution connected to governance controls for access, transformation, and release.

The outcomes I care about include expert agreement, calibrated uncertainty, unauthorized disclosure, mistaken approvals, unnecessary blocking, and legitimate task completion. Technical attribution, legal ownership, and permission to use an asset each require their own evidence and decision process.

AI governance: from policy to observed behavior

Archetypal is the company most directly connected to my AI governance research. Its work concerns policy reasoning, decision records, enforceable action boundaries, and the organization of governed agents. The research asks whether agents can maintain authorized behavior as instructions, incentives, memory, and operating conditions change.

The Core Eight framework connects authority, evidence, policy decisions, enforcement, adjudication, resources, recovery, and learning. Related work explores Rule Description Logic and policy-engine extensions, alongside proposed studies of persistent role identity, governed memory, and bounded feedback. The collaboration record includes an Archetypal Federal–NSA cooperative research and development agreement signed in November 2025.

At GW, I want to test these ideas with equally resourced comparison systems. The measures would include executed violations, useful task completion, human correction, repeated failure, recovery, and review burden. The current papers and architectures define hypotheses and study designs; the next step is rigorous experimental evaluation.

The technology and IP background I bring

The background spans several distinct bodies of work. Spyris contributes its platform, attribution extension, provenance research, and associated invention-development work. Archetypal contributes governance architectures, policy-engine research, decision-recording concepts, and related invention drafts. My subsequent research around NSA policy-engine technology includes reviewed Rule Description Logic extension work.

Spyris’s amended and restated NSA patent-license agreement, signed in June 2026, lists 14 distinct patent numbers. One is U.S. Patent 10,042,928, System and Method for Automated Reasoning with and Searching of Documents, which is particularly relevant to the policy-engine research. The wider schedule spans document and word relevance, similarity measurement, graph and streaming computation, security, image-source attribution, encryption, and reversible computation. The public patent records are linked below.

These NSA-developed patents form licensed background technology. Company-developed software and research, proposed derivative inventions, and new academic results have their own ownership and review histories. The reviewed invention drafts establish work in progress; filing and inventorship status must be established through the corresponding formal records.

For a university collaboration, I would bring a documented inventory of the relevant assets and agreements so that permitted research use, publication arrangements, and the treatment of new results can be defined at the outset. Specific access and sublicensing would follow the applicable agreements and approvals.

Future Works and DataBank: extending the core platform

My broader direction is to extend the Spyris core into settings where technology, data, and human contributions move across organizational boundaries. AI Labs Future Works is intended to carry the innovation workflow into technology evaluation, product development, venture creation, and commercialization. DataBank is a further direction for applying the same foundation to governed data and knowledge use.

These expansion directions create research questions of their own: can provenance remain intact as an idea becomes a product? Can permissions remain interpretable as assets are transformed or shared? What information helps reviewers make an appropriate decision at each handoff? I want to study these questions in scoped, observable workflows as the applications develop.

A foundation in scalable computing and human health

My earlier research provides another part of this story. Work with Soo-Yeon Ji explored cloud-scale processing and classification of gene-expression microarray data. Coauthored work in Nature Medicine examined individual and community-level COVID-19 mortality risk, while a Nature Genetics study examined the plasma proteome across European and African ancestry groups.

Alongside a review of CUDA, MapReduce, and Pthreads, these publications reflect a sustained interest in scalable analysis and the context needed to interpret evidence. My experience at the NSA, Databricks, and AI Squared added the practical challenge of bringing analytical systems into real workflows.

The academic collaboration I am looking for

I want to develop a research program at the intersection of trustworthy AI, human decision-making, governance, and technology provenance. A shared, unclassified research sandbox could connect my engineering background with faculty and student work on controlled experiments, expert evaluation, and appropriate reliance.

My immediate priorities are a longitudinal agent-governance study and an expert-grounded attribution-and-governance study. Connected directions include authorized AI security testing and coordination across distributed agents; federated AI would extend the work to data locality and participant authority.

The goal is a coherent body of publishable research that tests the ideas, measures their limits, and produces knowledge others can build on. That is the next chapter I want to pursue at GW.