Work With FERZ
Build the runtime authorization layer for AI.
The FERZ Professional Philosophy
Most teams treat the unpredictability of a model as a fixed constraint and build observability around it. FERZ takes a different position. The point of control is not the model. The point of control is the action the model attempts.
We govern that action at execution time through a non-bypassable enforcement boundary that returns one of three verdicts: ALLOW, DENY, or ABSTAIN. ABSTAIN blocks execution pending authorized human override. The model stays probabilistic. What it is permitted to do is determined, recorded, and verifiable.
This is distinct from identity and access management, which authorizes principals to reach resources, and from AI observability, which records behavior after the fact. Observability is not authorization.
Execution-time authorization, the three-verdict space, and the authorization artifact are defined in the FERZ research corpus. See On the Impossibility of Observability-Based Authorization and the full research corpus.
Areas of Interest
Formal rigor
We rely on formal methods, deterministic outcomes, and structural reasoning where the problem allows it. We separate what is proven from what is asserted, and we mark which is which.
Systems thinking
We build the authorization boundary as infrastructure, structurally independent of the system it governs, rather than as a feature added to an application. Governance is a property of the architecture.
Regulatory reality
We build for regulated environments where an organization must be able to show why an AI action was permitted, not only that it was logged.
Where We Build a Bench
We are building a network of qualified professionals across the disciplines that runtime AI authorization requires.
AI/ML Systems Engineering
Senior-level, project-based and full-time interest
Build the runtime infrastructure that evaluates AI actions against encoded policy at execution time and enforces the verdict fail-closed. Work spans the enforcement boundary, verifiable authorization artifacts, and the systems that keep authorization independent of the system it governs.
Key expertise areas: formal verification, constraint-based reasoning, distributed systems, Python, systems performance, enterprise authorization infrastructure.
Regulatory Compliance & Policy
Mid to senior-level advisory interest
Translate legal and regulatory frameworks into enforceable authorization policy. Work at the boundary between what a regulation requires and what an enforcement system can evaluate at runtime.
Key expertise areas: EU AI Act, FDA AI guidance, SEC algorithmic oversight, privacy law, government contracting, risk management, policy analysis.
Enterprise Consulting & Implementation
Senior consulting interest, large-organization engagement
Guide organizations through placing a runtime authorization boundary into existing AI systems. Lead implementations, align stakeholders, and deliver authorization outcomes that hold up under audit.
Key expertise areas: enterprise transformation, change management, AI governance consulting, stakeholder engagement, project management, financial services and healthcare domain knowledge.
Product Development & Architecture
Mid to senior-level product and engineering interest
Design the interfaces that make runtime authorization legible to the people who configure and review it: policy authoring, verdict review, and the authorization artifact as user-visible evidence.
Key expertise areas: product management, UX design, full-stack development, API design, enterprise software, technical product marketing, user research.
Computational Linguistics & Research
Research and advanced development, PhD preferred
Advance the formal treatment of meaning that lets policy be expressed precisely and governed across versions. Publish into the FERZ corpus.
Key expertise areas: computational linguistics, formal semantics, formal methods, natural language processing, academic publication, theoretical computer science.
Strategic Partnerships & Business Development
Senior business development and partnership interest
Build relationships with cloud providers, systems integrators, and enterprise software companies. Develop licensing and channel strategies that position FERZ as the authorization layer within larger stacks.
Key expertise areas: enterprise sales, partnership development, IP licensing, strategic alliances, channel management, competitive intelligence, market analysis.
Important: Expression of Interest, Not a Job Posting
This is not a job posting. FERZ is building a qualified network for future opportunities that funding milestones and client engagements may create. An expression of interest does not guarantee consideration, an interview, or employment.
Contingent opportunities. Potential engagements are contingent on funding milestones, executed client contracts, security clearance requirements for government work, and completion of the FERZ evaluation process.
Selective process. FERZ evaluates at its discretion and may decline an expression of interest without explanation. We look for technical capability together with alignment to enforcement-first AI governance.
Expression of Interest Process
Submit interest
Complete the form with your background, areas of expertise, and your perspective on enforcement-first AI governance.
Initial evaluation
FERZ reviews submissions for technical qualification, domain depth, and fit. Some qualified respondents may be added to the network.
Network engagement
Members receive periodic updates on FERZ progress and early notice when specific opportunities open. The aim is a durable professional relationship, not only filling roles.
Opportunity activation
When funding or a client engagement creates a specific role, FERZ may contact aligned members of the network about a formal process.
A Personal Note from Edward Meyman, Founder
FERZ exists because the unpredictability of a model is not a reason to leave its actions ungoverned. Many people read the language in an LLM as statistical relationships between tokens. I read a structure that can be governed with precision. The model may remain probabilistic, but the actions it attempts should pass through a runtime authorization boundary before they take effect.
If you see formal structure where others see noise, if you value a verdict that can be independently reconstructed over a best-effort explanation, and if you are willing to hold a contrarian position long enough to build it, I want to hear from you.
We are building the authorization layer that lets an organization show that an AI action was permitted under binding policy before execution and remains provable after the fact. That is the work.
Express Your Interest
Join the professionals working on enforcement-first AI governance.
This is a talent network, not a job posting. Submitting your details lets us reach out if an aligned opportunity opens. We are building a community of people who share the position that AI actions should be authorized, not only observed.
We Welcome Qualified Professionals
If you do rigorous work and you hold that AI governance is an enforcement problem, we want to connect. FERZ is selective because the category is new and the boundary must be defensible before execution, not explained afterward.
FERZ uses the information you submit to evaluate and respond to your expression of interest. See our Privacy Policy for details. By submitting this form, you acknowledge that this is an expression of interest only and does not create any employment relationship or guarantee of future opportunity. FERZ evaluates and may decline at its sole discretion.