AI Governance Architect

NorthHill Technology Resources has an urgent need for an AI Governance Architect to support a financial client in Mclean, VA. This is a 9 - 12 month contract engagement. It is a hybrid role, with 3 days onsite and 2 remote per week. Position Summary Lead the design of the bank's AI orchestration and guardrails layer so approved copilots and agents operate with policy enforcement, routing, logging, traceability, and human review controls. This role converts the current AI Policy control set into a usable technical blueprint that internal staff can implement, operate, and defend to Audit and examiners. Core Responsibilities Design the AI control plane for policy enforcement, workflow routing, logging, exception handling, and release management across approved AI tools and workflows. Define reusable approval patterns for use-case intake, sensitivity-based review, AI Inventory registration, connector enablement, plugin / MCP approval, and escalation to IT, Compliance, Legal, Security, and Risk. Establish technical standards for model access, fallback behavior, kill switches, approval checkpoints, and supportable handoff from pilot to production. Translate the Human Verification tiers in policy into workflow patterns, including full review, spot-check review, and workflow acknowledgment where appropriate. Design evidence capture so use-case approvals, exception decisions, logging, and monitoring artifacts flow into the governance repository with minimal manual effort. Define architecture patterns for desktop agents, scheduled tasks, plugins, skills, and custom MCP services so new capabilities inherit controls instead of bypassing them. Partner with IT and Security to specify identity, access, and environment controls for approved enterprise AI platforms, including Claude Cowork, Claude Code, Microsoft tooling, and the SharePoint-based Presidio and Protect masking path where applicable. Establish release-readiness criteria and validation checkpoints for logging, evidence capture, fallback handling, and control verification before governed AI workflows go live. Create operational runbooks for exception handling, incident response, emergency disablement, and audit support so control issues can be triaged consistently. Guide implementation choices for application and integration engineers and produce the design package, runbooks, and standards needed for long‑term internal ownership. Control Requirements Define the standard for AI workflow logging, including metadata, connector activity, desktop‑agent or scheduled‑task activity where enabled, and evidence retention expectations. Specify how the platform will enforce role‑based access, approved prompt patterns, restricted capabilities, and escalation to human review for higher‑risk actions. Set governance patterns for approved connectors, plugins, custom MCP servers, skills, standing prompts, and SharePoint‑based masking and unmasking services so new capabilities cannot bypass the approval workflow. Design incident and exception pathways for verification failures, output errors, control breaches, and emergency disablement of AI features. Document integration points with AI Inventory maintenance, quarterly reporting inputs, and audit or regulator evidence requests. Required Qualifications 8+ years in enterprise architecture, platform engineering, security engineering, workflow orchestration, or control‑plane design. Demonstrated experience designing systems in regulated or highly controlled environments where approvals, logging, and segregation of duties matter. Strong background in identity and access control, observability, audit trails, API orchestration, event‑driven workflows, and production release governance. Ability to map policy or risk requirements into pragmatic technical standards and communicate them clearly to both engineers and business stakeholders. Experience creating implementation blueprints, decision records, and handoff documentation that internal teams can sustain after the initial design period. Preferred Qualifications and Skills Experience with enterprise AI governance, policy enforcement, model allow‑listing, or responsible AI implementation in Microsoft or comparable ecosystems. Familiarity with Microsoft 365, Azure, Copilot Studio, Power Platform, SharePoint, or workflow tooling commonly used in controlled enterprise environments. Financial services, fintech, broker‑dealer, insurance, or other regulated‑industry experience with audit and compliance exposure. Exposure to records‑retention, incident‑response, and evidence‑management requirements that influence platform design. Familiarity with Claude Cowork or similar enterprise AI environments used for governed document review, analysis, and human‑in‑the‑loop workflows. Familiarity with Claude Code or comparable AI‑assisted engineering tools used for secure prototyping, implementation acceleration, or control‑design support. #J-18808-Ljbffr