Researchers
VETPOC Labs is Virtual Enterprise Trust, Policy, Oversight & Conscience, the research program Arun V. Kadambi leads on institutions for AI-generated software change.
Research lead
Research lead of VETPOC Labs, and a PhD student in Information Science at the University of North Texas.
He studies how an enterprise can govern software changes that AI agents propose: which acts may be allowed, who has authority to make an exception, and which decisions should bind the next case. The two instruments in the program are Enterprise Conscience, a policy gate, and Digital Judiciary, a framework for contested decisions.
The dissertation named in the program materials is The Digital Judiciary: A Framework for Governing AI-Produced Code Changes and AI-Made Decisions. A related preprint with Venkat Kuppuswamy is The Digital Judiciary: Binding Precedent for Automated Compliance Decisions.
Experience
Roles below follow the biography Arun supplied for this page. His public LinkedIn profile is linkedin.com/in/arunvj. That profile shows his name and Broadcom employment, including a headline as Global R&D Director and business-unit executive for an AI operations and observability portfolio. It is not a source for every role, date, or degree here. Parts of the public profile are not fully visible.
Broadcom · current
He leads global teams on predictive intelligence and agentic automation for enterprise customers. Observability is part of the portfolio named in his public LinkedIn headline. Workload automation is named in the biography he supplied. The biography does not fix a single list of regions, so none is stated here.
Prior executive role
The biography describes this role as a large-scale modernization of mainframe software, from manual operation toward automated self-service. It does not name the employer, so this page does not assign one.
CA Technologies
He led development and launch of an autonomous anomaly-detection platform for mainframe environments.
Oxygen Finance
The biography names this as an earlier leadership post in product incubation.
E2open
At E2open he worked on a data-driven supply-chain control-tower cloud service and directed a global cloud collaboration center.
Education
Degrees and honors are as given in the biography he supplied. This page does not claim they were checked on LinkedIn. The doctoral enrollment is the affiliation already stated in the program materials.
University of North Texas · doctoral study
Doctoral affiliation, separate from VETPOC Labs. Not a university laboratory and not an institutional endorsement.
The University of Texas at Austin
The University of Texas at Dallas
The biography notes Academic Excellence and Class Leadership Awards.
College of Engineering, Guindy
Collaborators
These are the collaborators already named on the program site. This page does not add biographies, offices, or contact addresses for them, and it does not add other researchers.
Roles are as stated in the November 2026 program site.
The program
Which agent actions may an enterprise ever allow, and for what reasons? Who has authority to grant exceptions? Which decisions should bind the next case, and what does that choice do to the record?
Formal models of precedent and permission, working software, preregistered simulations with computed ground truth, analysis of legal corpora, and practitioner surveys. Several of those pieces are still manuscripts. The evidence page separates what is calibration from what is pending.
The program site says the questions, frameworks, designs, and decisions are the authors’, and that AI tools assist with drafting, literature search, and code under the authors’ direction, with that use declared in each paper.
Contact
Practitioners who govern AI-assisted software delivery, legal scholars working on precedent and contestability, and researchers interested in replication or extension can write. There is no form on this site, no mailing list, and no analytics. The research address is the one already published in the program materials.
A conversation is not a deployment. The prototypes are not offered as a control for a production pipeline.