I work on AI policy
at the layer where
promises get tested.

An interrupted time series analysis of US federal contracting data examining whether the 2024 Office of Management and Budget (OMB) AI procurement memoranda measurably shifted agency-level AI adoption.

Identifies AI-related contract actions across roughly 7,200 agency-months and tests for level shifts at the M-24-10 (March 2024) and M-24-18 (September 2024) intervention points. Distinguishes between extensive-margin (frequency) and intensive-margin (volume) responses — a distinction with direct implications for how administrative AI policy is read as a regulatory instrument. Full chart and methodology →

An argument that the persistent failure of public-sector AI pilots to scale into deployment is not a technology problem or a funding problem — it is a structural feature of how the pilot apparatus itself is institutionally built. Pilots succeed because they are exempted from the procurement, governance, and accountability structures that deployment would have to operate within.

The same exemptions that make pilots succeed are the conditions that prevent them from progressing. Stagnation is the system working precisely as the institutional environment has forced it to.

Argues that the EU AI Act draws no coherent conceptual line between harmful and productive information asymmetry — Article 5 prohibits certain forms of opacity outright, while Articles 57–63 institutionalise opacity as a precondition for innovation.

The Act's boundary between the two is not principled but procedural and distributive: it calibrates who benefits from informational asymmetry and on what terms, rather than resolving the underlying tension.

Contributed to the design and adoption of the County's first governance framework for frontier AI models, establishing institutional procedures for evaluating, piloting, and scaling emerging technologies across departments serving over 500,000 residents.

Led qualitative stakeholder interviews with department leads across government divisions and synthesised findings to shape the framework's evaluation, pilot, and scale-up procedures.

Co-designed an 8-criterion structured assessment framework for evaluating frontier AI tools against procurement requirements, workflow constraints, governance risk, and political context. Combined qualitative evaluation with light quantitative scoring to reduce evaluator bias and produce a consistent evidence base for procurement decisions.

Used to assess frontier AI tools for County adoption; outputs informed leadership's technology decisions and were incorporated into the broader governance framework.

Authored five pilot proposals for the project management team evaluating frontier AI tools for departmental adoption. Each proposal assessed implementation feasibility, governance risk, and operational impact — translating abstract technology guidance into concrete pilot designs that the County could actually run.

Mixed-methods study using qualitative and archival analysis of procurement documentation, government reports, and organisational case studies to trace how procurement failures upstream of the 2013 Healthcare.gov rollout determined delivery outcomes.

The argument: the launch crisis was a procurement crisis before it was a software crisis. Sets up the procurement-as-governance frame extended in the MSc thesis.

Managed a 3-month research project from scoping through publication, delivering a 15,000-word policy report on AI and democracy promotion. Incorporated a literature review of 50+ sources, stakeholder insights, and a policy memo with recommendations now used to inform organisational strategy.

Developed a working definition of AI for the democracy support sector and translated it into a briefing deck presented to WFD staff and external partners, packaging complex policy research for varied audiences.

Eight-week fellowship project on AI's impact on democratic practices and elections — synthesising primary sourcing, case studies, stakeholder interviews, and quantitative data collection. The report deliberately resists the genre's pull toward speculation; the goal was a usable baseline rather than scenario-building.

A 5,000-word essay produced under tutorial supervision at Oxford, examining how states — particularly those with limited means or resources — should respond to challenges to their authority from digital platforms. Treats sovereignty as a multidimensional resource rather than a binary status, and traces what kinds of state responses are actually viable across legal, physical, and political dimensions.

Selected by peers to draft the youth statement for the 2023 Nuclear Non-Proliferation Treaty Preparatory Committee, and delivered the address on the role of information and communications technologies in security policy. Briefed parliamentarians and members of Congress on the AI–nuclear nexus throughout 2023.

Prepared speech materials for civil society partners and UN representatives, helping communicate policy positions on AI governance, emerging technologies, and international security in clear, public-facing language.

Researched the AI–nuclear nexus and produced internal-use policy guidance on organisational use of generative AI for the Nuclear Age Peace Foundation. The guidance covered acceptable use cases, governance considerations, and the boundary between research and operational deployment in a small-organisation context.

Spearheaded internal discussions with NAPF staff about implementing the guidance in day-to-day work.

Drafted executive appointment orders and governmental memos at the Office of Boards and Commissions; managed candidate pipeline across 12 active board vacancies, coordinating with appointees and agency staff on state-level operations.