About
Mary Ann Belarmino, Certified AIGP
Founder & Principal
Most AI governance programs are built by two groups that rarely speak the same language. Legal, risk, and policy teams know what the rules require. Engineering teams know how the systems actually work. The space between them is where governance breaks down: policies that cannot be tested, controls that exist on paper but not in the pipeline, and audit evidence that cannot be traced back to a real system.
Mary Ann works in that space. Her practice brings together AI governance, quality systems engineering, technical product management, and strategy and implementation, because in practice they are one problem: governance without product context stalls, and quality without governance cannot show that a system is safe to ship. She brings more than 25 years across embedded engineering, enterprise software, quality engineering, and organizational transformation at MathWorks, Pegasystems, Western Digital, and IntelyCare, built on a foundation in applied physics and published research.
For enterprises, she translates regulatory and policy requirements into controls, test plans, and evidence that engineering teams can build and auditors can verify. For government, nonprofit, and standards bodies, she brings the engineering reality of how AI systems are built, tested, and operated, so guidance can be put into practice.
What sets her work apart is that she does not apply these skills one at a time. On any engagement, she looks at the same problem as an engineer, a product manager, a quality leader, and a change leader at once, and traces how each part affects the others. Her quality transformations show how this works. She starts on the product side, helping teams define clear user stories and scope, so engineering and QA receive well-scoped work. That leads to faster releases, less rework, and fewer incidents. When incidents do happen, she leads root cause analysis and characterizes the failures, then feeds what was learned back to product and engineering so the whole cycle improves. AI governance needs the same closed loop, from intended use through testing, release, and monitoring.
She does not replace your counsel, your risk team, or your engineers. She connects them, so your AI governance works in practice and not only on paper.
What she brings
Open any area for detail.
Technology depth, from hardware to enterprise platforms
Mary Ann started as a software engineer on embedded systems and digital signal processing, working close to the hardware. She went on to build developer tooling and lead quality engineering, production support, IT, and transformation work in enterprise software at MathWorks, Pegasystems, Western Digital, and IntelyCare.
At MathWorks, she built the embedded development tooling (ToolchainInfo and the integrated build process) that became the foundation of embedded development at the company. She also led the migration and refactoring of a 20-year-old legacy subsystem to a new architecture that became the foundation of three critical Coder products. She is a named inventor on U.S. Patent 7,490,032.
She contributed a working MATLAB/Simulink example that connects to a DSP chip through Real-Time Data Exchange (RTDX) to Rulph Chassaing’s Digital Signal Processing and Applications with the C6713 and C6416 DSK, and served as a technical reviewer for Head First Design Patterns (O’Reilly).
Because she understands systems from low-level hardware to enterprise platforms, she can follow an AI risk through every layer it touches.
Technical product management
At MathWorks, Mary Ann held a combined role as lead software and embedded engineer and technical product manager, shaping the direction of the embedded development tools she also built. At Pegasystems, before becoming global head of Platform quality, she was a Senior Product Manager who delivered developer tools capabilities to Pegasystems business clients.
This experience lets her see governance from the product side: how requirements, roadmaps, and release decisions are made, and where governance controls need to fit so they support delivery instead of blocking it.
Mathematics and research foundation
Mary Ann earned her BS in Applied Physics at the National Institute of Physics, University of the Philippines Diliman, where she built the mathematical foundation that underlies modern machine learning, including signal processing, algorithms, and optimization.
As lead author, she published her undergraduate research in Applied Optics (1998), a peer-reviewed journal of the Optical Society of America (published as Mary Ann Nazario). The work showed how to recover signal information lost during sampling, even in the presence of noise, using a constraint-based mathematical method (the minimum-negativity constraint).
She went on to complete graduate coursework in electrical engineering at UP Diliman, including electives in neural networks, robotics, VLSI design, audio engineering, and algorithms, and carried out a full graduate research project on Hidden Markov Models, from proposal through written thesis and results.
This grounding lets her question how a model was built and evaluated, not just whether it passed a checklist.
Quality engineering and operations
Years of leading quality engineering and production support gave Mary Ann an eye for detail and a firm rule: a control is only real if you can test it. Production support taught her how systems fail in practice, not in the design document.
At IntelyCare, she was head of Quality Engineering, Performance, Production Support and Incident Management, Change Management, and IT. There she created a User Acceptance Testing group that validated features against user requirements, tested complete end-to-end user journeys to confirm customers could reach their goals, and evaluated the user experience for usability.
She set the direction for redesigning Production Support and Change Management and worked hands-on with the managers who led those groups to define the methods that made the changes possible. She streamlined the intake of support tickets and incidents so that customer pain points became usable insight. Issues reached product and engineering teams with the information needed to act on them, such as root cause and severity.
In Change Management, she streamlined how changes were released to production, so that defined criteria were met before any release or update went out, and the right downstream teams were informed, especially when a change directly affected customers.
All three practices carry directly into AI governance. Testing a system against its intended use and real user journeys is how AI systems should be evaluated. Release criteria and stakeholder notice are the same controls organizations need before an AI model or update reaches users. And a working incident feedback loop is what AI regulations increasingly expect through post-deployment monitoring and incident reporting. She applies the same discipline to AI governance, turning policy statements into controls with clear test criteria and evidence that can be traced back to the system they govern.
Organizational change leadership
Governance is an organizational change problem, not only a technical one. In every role she has held, Mary Ann has worked to change mindset and culture, not only process and tools. She has led enterprise quality transformations built on the same foundations that AI governance requires: policy, strategic initiatives, people, training, process, and technology.
Those transformations moved quality checks earlier in development (“shift left”), changed how teams thought about their own responsibility for quality, and built habits of continuous learning and improvement. She has done this work in engineering organizations ranging from about 20 engineers to about 1,000, including as global head of Platform quality at Pegasystems.
Along the way, she turned teams that were seen as cost centers into multipliers: releases became faster and more frequent, defects, incidents, and rework dropped, and the capacity this freed up allowed new projects to move forward.
Systems thinking for connected and agentic AI
Her physics and engineering training, along with executive study in systems thinking at MIT, gives Mary Ann a systems view of how risk moves and compounds across connected parts.
That view matters more as organizations deploy AI agents that act on each other’s outputs, where a failure in one component can spread in ways no single team owns.
Teaching and technical communication
Mary Ann taught calculus, physics, electrical engineering fundamentals, and electromagnetism at the university level, in the Electrical Engineering department of UP Diliman and at the University of Asia and the Pacific. She has also spoken at quality and testing summits.
Teaching gave her the ability to explain technical risk clearly to boards, regulators, and non-technical leaders.
Executive education and credentials
Her executive education includes business analytics at Wharton; data leadership and systems thinking at MIT; strategic financial analysis at Harvard; and execution strategy for public leadership at the Harvard Kennedy School. She holds the IAPP’s AIGP credential.
Her training in public leadership and strategic financial analysis helps her frame governance in terms of mission, budget, and execution, not only compliance.
Original AI governance work
Mary Ann develops original frameworks, protocols, and diagnostic tools for gaps she sees in current AI governance practice, including data protection, governance maturity, regulatory obligations across jurisdictions, and multi-agent AI risk. Published work is on the Frameworks page, and overviews of additional methods are available on request.