FJ O’Shea

What FJ Covers

FJ writes for non-technical professionals thinking about AI at work — whether they are curious about AI and not yet using it, worried or skeptical about its impact on their career, excited about its potential, already using AI at work and choosing what to use next, following AI as a topic out of intellectual or cultural interest, or leading AI adoption for a team or organization. He covers anyone in an office-based role that involves reading, writing, analyzing, managing, advising, or selling, deciding which AI tools and courses are worth their time and money. His coverage spans AI evaluation methodology, AI workflow design, AI courses and certifications, AI for specific white-collar verticals, and the career impact of AI on professional work.

The goal is decision quality: help readers understand what is worth their time, what is worth their money, and what should be skipped.

Background and Experience

FJ’s operational discipline comes from ten years in financial-services back- and middle-office work. At State Street (2012–2015) he progressed from Securities Valuations into the OTC Derivatives Centre of Excellence, ending as Senior Associate. At MUFG Investor Services (2016–2022) he held progressive middle-office and BPO operations roles in Dublin and Limassol, ending as Director, Head of Middle Office Operations, Cyprus.

The work covered trade processing, valuations, client onboarding, complex reconciliations, automation projects, and operational change at scale. Controls, SLAs, auditability, escalation paths, and exception handling defined service quality. That decade of operational rigor is what FJ now applies to AI systems.

Research and AI Evaluation Work

FJ co-authored peer-reviewed research applying graph-based machine learning to operational workflows: Frangos & O’Shea (2022), A graph-based approach to client relationship management in fund administration, Machine Learning with Applications, 10, 100433. DOI: 10.1016/j.mlwa.2022.100433. His contribution was the domain methodology, classification framework, and ground-truth labelling of 8,552 operational emails. The technical machine learning implementation was led by his co-author.

Since 2024 FJ has worked as an independent AI evaluation contractor on confidential evaluation programs for frontier large language models. The work covers:

• Factuality verification with multi-tier accuracy and YMYL severity assessment
• Sentence-level groundedness and source-attribution review
• Multi-turn dialogue evaluation and generative editing
• Voice and mobile assistant evaluation including tool-call verification
• Multi-hop prompt engineering for adversarial error elicitation
• Visual reasoning and multimodal output evaluation
• Adversarial multi-file agent evaluation across accounting, finance, and healthcare professional task domains

FJ was quoted in Lifewire’s August 2023 piece on AI voice cloning, discussing both practical workflow uses and the deception risks of cloned voices.

How FJ Reviews Products and Courses

Aiifi reviews are decision-first. FJ tests AI tools and evaluates AI courses against the criteria that matter for non-technical professional buyers: claim accuracy, real-world workflow fit, time-to-value, and total cost. Where first-hand testing applies, completion timestamps, methodology notes, and original screenshots are included on the page. Source material is checked against vendor documentation and provider pages. Comparison is preferred over standalone scoring.

Aiifi’s editorial standards cover the full review process: how products and courses are selected, how completion or first-hand testing is verified, what disqualifies a product from recommendation, when AI assists in research or drafting, and how affiliate relationships are disclosed. Read Aiifi’s full editorial policy and review methodology. See also the affiliate disclosure and corrections process. For questions or correction requests, get in touch.