Founder and CEO, Forge Orbital

Joseph Varga

Building independent AI decision accountability for high-consequence enterprise and federal systems.

The system making the decision cannot be the only system explaining why the decision was justified.

LinkedIn X Email Read the work

The work

Accountability has to survive the system being wrong.

Joseph founded Forge Orbital around a structural problem in enterprise AI: most governance evidence is written by the same organization and technical stack whose decision is being reviewed. That is useful for operations. It is not independent proof for a regulator, board, auditor, insurer, customer, or program office.

Forge turns the supplied evidence, encoded policy result, uncertainty, action menu, and human authority around a selected high-consequence decision into a defensible proof trail. The customer retains the policy and execution authority. Forge supplies a deterministic disposition and a verification path that can be checked outside the original workflow.

Joseph built the Forge decision engine and leads its enterprise and federal work across AI agents, financial services, insurance, critical infrastructure, and mission systems. Before Forge, he spent three years as an executive in defense technology. He earned his BA from Wabash College.

The category in plain English.

Independent AI decision accountability

A checkable basis for why a high-consequence AI or automated action cleared, held, or escalated under the evidence and policy supplied at decision time.

Problem

The self-grading gap

The organization is asking an outside reviewer to trust evidence written by the same system being questioned.

Product

AI decision proof layer

A deterministic layer beside the workflow. It records the decision basis without becoming the customer’s decision authority.

Output

Defensible proof trail

The evidence, rule, uncertainty, disposition, human boundary, integrity check, and replay basis for one selected action.

Current areas of focus.

01 / Enterprise agents

Accountability at the action boundary

Give a reviewer a checkable reason for a selected agent action without pretending a prompt, dashboard, or self-authored log is the control.

02 / Financial services

Model and decision governance

Preserve the basis for model releases, trading controls, exceptions, approvals, and AI-assisted financial decisions.

03 / Insurance

Evidence for underwriting and claims

Map messy source material into bounded evidence and preserve what supported the action, what remained uncertain, and who kept authority.

Selected writing.

Category definition

Bring the decision that cannot be defended from memory.

The first conversation is one workflow, one outside reviewer, one action menu, and the evidence currently scattered across systems.

Contact Joseph See how Forge works