Janet Xiu Shi - AI Quality Evaluation Advisor and AI Educator, Ireland Janet Xiu Shi

AI adoption that stands up to scrutiny.

I'm Janet Xiu Shi, an AI Quality Evaluation Advisor and AI Educator based in Ireland, serving founders, CEOs, and senior leaders internationally.

  • I solve business problems. AI is a tool, not the default answer. I diagnose what's actually broken or missing, map it to the right technology, AI or not, and make sure what gets built or adopted actually fixes it.
  • Results you can measure, not activity you can describe. 32% faster development cycles and zero SLA breaches at Workday. 52% inventory cost reduction re-engineering Huawei's European supply chain. If it didn't move a real number, it isn't finished.
  • Independent judgment, not a vendor's pitch. I tell you when AI isn't the answer, when what's live isn't working, and what to do instead, before you scale a mistake or walk away from something that would have worked.
  • Bilingual Mandarin/English, cross-culture. I work directly with Chinese-speaking and English-speaking stakeholders alike, no interpreter, no lost nuance, in either direction.
    全景式数智化顾问，跨界赋能者让战略落地，让智能生效。
    懂大厂，也懂创业；懂中国，也懂西方。这样的AI判断，才靠谱。
See how to work with me ↓
Agent Boutique
Maya by Janet Xiu Shi™ - AI-generated avatar, edited for production. Disclosed as AI Generated in line with the European Commission's optional icon-labelling guidance for AI-generated content Maya™ Live digital twin Ming - AI quality agent working in the background. Disclosed as AI Generated in line with the European Commission's optional icon-labelling guidance for AI-generated content Ming Working Behind the Scenes Maggie - Data Powered Marketing, Brand and PR agent working in the background. Disclosed as AI Generated in line with the European Commission's optional icon-labelling guidance for AI-generated content Maggie Working Behind the Scenes

Institutional credibility

Huawei Technologies Workday Connectd Avencia Consulting Woodies CH Interior&Fashion Dunnes Lidl GmbH
Harvard University CS50 Anthropic Academy IADT SOLAS IDA Ireland Enterprise Ireland ETB Hartley People Training GDG Ireland
17
global locations
scaled from Dublin
2,000+
community members
GDG Ireland
32%
dev cycle reduction
Workday enterprise
52%
inventory cost reduction
Huawei supply chain
20+
years cross-industry
enterprise experience

Agent Boutique by Janet™

The model isn't the edge. Your data, your domain knowledge, and your business insight are, and the right technology built around them is what supercharges a team.

Agent Boutique builds purpose-driven agents for your business, trained on your own data and workflows, not off-the-shelf demo copy. Maya, just below, is the flagship - live, testable, and running the same standard you'd get.

See the full boutique →

Meet Maya

My live AI digital twin, and a working example of the systems I build, evaluate, and govern. Not a mockup.

Maya by Janet Xiu Shi - AI-generated avatar of her AI digital twin, edited for production. Labelled AI Generated, disclosing upfront that Maya is an AI system in line with the European Commission's optional icon-labelling guidance for AI-generated content
Agent Boutique · The flagship you can test

Maya is my live AI digital twin, born 1 Feb 2026 and live ever since. What is running today is her second generation, rebuilt on AWS in July 2026, so her consent, retention and transparency behaviour sits under an automated test suite rather than under my word for it. As your AI Advisor, she diagnoses where AI actually fits in your business and how to build guardrails and evaluation frameworks from day one. As your AI Educator, she brings pedagogical practice from Harvard CS50, IADT, and QQI AI Micro-Qualifications, challenging learners to build strong fundamentals first, then use AI to deepen learning through iteration, verification, and responsible use.

She's also a working example of the governance I advise on, not a slide about it: Chat content processed in AWS's EU infrastructure and stored in the EEA; consent enforced before any conversation is logged; automated guardrails; and a defined 90-day retention and erasure policy under GDPR. Maya and I believe the best AI is experienced, not advertised. It works quietly underneath daily life, holds up under scrutiny, and improves quality of life rather than chasing attention.

✨ Talk to Maya ✨

Meet Ming

Quality and governance, already running on every project before it ships.

Ming - AI quality and governance agent. Labelled AI Generated, disclosing upfront that this image is an AI system output in line with the European Commission's optional icon-labelling guidance for AI-generated content
Agent Boutique · Checks every build

Ming is my quality and governance agent. Nothing I ship reaches a client, a classroom, or this page until she has checked it: that the tests actually ran, that the accessibility holds, that the security review happened, and that every factual claim traces back to something real. She holds the sign-off, and she uses it.

That standard is not aspirational. It is the same 6-point definition of done my own live systems have to pass, and Ming fails my work against it when it does not pass. If you want to know how I will treat your AI system, look at how I treat mine.

Working Behind the Scenes

Meet Maggie

Positioning that survives a fact-check.

Maggie - Data Powered Marketing, Brand and PR agent. Labelled AI Generated, disclosing upfront that this image is an AI system output in line with the European Commission's optional icon-labelling guidance for AI-generated content
Agent Boutique · Checks every claim

Maggie is my Data Powered Marketing, Brand and PR agent. Most AI marketing fails in the same place: it sounds confident and cannot be checked. Maggie is built the other way round. Every recommendation she makes names the data behind it, and every claim about my work is verified against the live system before it reaches a page.

The first brand she audits is my own. She tracks where my job title, my regulatory dates, and my offer names drift apart across the sites I own, and she flags anything that would not hold up if a regulator or a journalist checked it. She is also building the layer underneath: who engaged, with what, and what the honest next conversation is. Attention is only worth something if it has somewhere useful to go.

That is also why you can take the rest of this page at face value. Every claim on it was checked against the system it describes, and the ones that could not be checked were cut.

Working Behind the Scenes

How to Work With Me

Three ways to start, each scoped to a decision you need to make -
not a menu of everything I know how to do.

For CEOs, founders & senior leaders

AI Decision & Readiness Sprint

"What should we actually do with AI, and are we ready to do it?"

  • Independent read on your current AI use, risk, and readiness
  • Priority-ranked use cases: what to pursue, what to pause
  • One recommended pilot, with an owner and a 90-day plan
  • A leadership decision briefing you can act on immediately

10-15 working days · fixed fee, scoped on a call

Start Here
For AI product, SaaS & agent teams

AI Product & Agent Quality Review

"Is what we've built actually reliable, safe, and ready to ship?"

  • Independent evaluation of outputs, failure modes, and edge cases
  • Data, access, and human-oversight risk review
  • A quality and evaluation framework you keep and reuse, anchored in AI-ARM (Accuracy, Relevance, Meaning)
  • A clear go / no-go / fix-first recommendation

2-4 weeks · fixed fee, scoped on a call

Start Here
For L&D, education & public sector

Workforce AI Capability Pilot

"How do we turn AI training into a habit that sticks, not just a course?"

  • Role-based learning paths tied to real workflows, not generic content
  • A pilot cohort that applies AI to an actual piece of their job
  • Before/after evidence of what changed, for leadership
  • A recommendation on what's worth scaling next

Scoped to cohort size · quote after a short call

Start Here

Every engagement ends with a clear recommendation on the next decision - including when the answer is "don't proceed."

The Track Record Behind It

The cross-domain pattern recognition that makes the AI work credible was built across 20 years of real operational complexity.

Foundation · 20+ years, now powering AI

Workday

Senior QA Engineer · 7.5 years

Enterprise-scale quality discipline across Financials, Supply Chain, and AI-assisted Expense workflows.

⚡ 32% cycle reduction
🏆 First VIBE Award

Huawei

Supply Chain Manager

Led Western Europe supply chain for telecom infrastructure across 15+ global ERP systems.

📉 52% inventory cost reduction

Retail Commercial Branding Leadership

Woodies · Dunnes · Lidl, 2002-2016

14 years building commercial and operational foundations across procurement and retail leadership.

🌍 €30M+ portfolio

Now · AI Advisory & Education

SOLAS & ETB

AI Programme Instructor & QQI Assessor

Appointed under SOLAS's national programme, delivered regionally through the Education and Training Board (ETB). Holds the core assessment role, grading learner work under QQI's rubric, peer-reviewed by a second assessor.

🎓 State-recognised qualification

IADT Lecturer

Higher Education, Ireland · QQI Level 8

Drawing on 20+ years across supply chain, SaaS, FinTech, and EdTech, teaching QQI Level 8 AI modules directly.

🎓 Higher education

AI Advisory

Connectd (London) · Avencia Consulting

Fractional board advisory and AI solution advisory for early-stage SaaS and growth companies.

🤖 AI solution advisory

Proof

Let the facts speak.

Janet Xiu Shi - Workday VIBE Global Impact Award

Workday VIBE Global Impact Award

Founded and scaled Workday's Employee Belonging Council from Dublin to 17 global locations. Won Workday's first-ever Product & Technology VIBE Global Impact Award.

🏆 First-ever award
🌍 17 locations
📈 30% YoY growth
View award ↗
SOLAS state-funded AI programme

SOLAS & ETB: State-Funded AI Educator

Appointed under SOLAS's national programme, delivered regionally through the Education and Training Board (ETB), to deliver and assess 2 national AI workforce programmes under QQI's own curriculum and rubric. One of a small group of practitioners in Ireland trusted with both instruction and assessment at this level.

🇮🇪 State-funded, QQI accredited
🎓 National programme
🤝 10 partner bodies
View programme ↗
Janet Xiu Shi - Lakera 2025 GenAI Security Readiness Report

Recognised Contributor, Lakera (2025)

Named a recognised contributor, Appendix C, to Lakera's 2025 GenAI Security Readiness Report for prompt injection and LLM security testing, ahead of Lakera's 2026 acquisition by Check Point.

🔐 AI security research
📄 Appendix C, 2025
Janet Xiu Shi - Harvard CS50 Professional Educator

Harvard CS50 Professional Educator

Harvard-certified Professional Educator for CS50 since 2021. Brings Harvard's foundations-first pedagogy into AI curriculum at IADT, SOLAS, and GDG Ireland.

🎓 Harvard CS50
Since 2021 to Present
The Proof Layer - real AWS and Amazon Bedrock builds by Janet Xiu Shi

Live AWS & Bedrock Builds

A dedicated proof layer of production AWS and Amazon Bedrock architecture, including Maya's multi-model pipeline, EU-region inference, least-privilege IAM, and real alarming. Built and operated, not described.

☁️ Production AWS
🤖 Amazon Bedrock
🔐 Governance in practice
View the proof layer ↗
Janet Xiu Shi - Champion Team, Founder Innovation Challenge 2024

Champion Team, Founder Innovation Challenge

Designed a growth and brand strategy for LandyCoo at Malaysia's Health & Beauty Innovation Conference. Champion team, Kuala Lumpur.

🏆 Champion team
Oct 2024

My Thoughts

Original frameworks built from two decades of operational and technical experience.
Not borrowed theory - tested discipline.

Evaluation Framework

AI-ARM: Accuracy, Relevance, Meaning

My anchor framework for assessing AI output quality. An output earns trust only when it clears all 3 bars: accurate (states what is true), relevant (answers what was actually asked, in context), and meaningful (changes a real decision or outcome). Most AI failures ship because only the first bar was checked.

Framework

5D Solutions Assurance

Discover → Design → Develop → Deploy → Deliver. A validation methodology for complex AI integrations. The Deliver stage is not publication - it is adoption measurement and feedback loop closure.

Quality System

Continuous P&R Quality Management (Calibrated to Context)

There is no universal Precision and Recall bar, and claiming one is itself a red flag. The right threshold is calibrated to what a false positive or false negative actually costs in your context, then held through continuous monitoring and recalibration, not a one-time certification.

Teaching Philosophy

Foundations First - Delay AI Use

Build core technical literacy before introducing AI-assisted workflows. Students and teams who understand how something works can evaluate, debug, and challenge AI outputs. Those who skip foundations become dependent on outputs they cannot verify.

Teaching Framework

ALE Scale: Applied Learning Evidence

Grading AI-era coursework on output quality alone is meaningless once AI can polish anyone's work. My ALE (Applied Learning Evidence) scale grades what's actually observable: whether a learner performs independently, needs guided support, shows only partial understanding, or hasn't shown evidence yet. Four honest categories, not a pass/fail guess.

Adoption Model

AI Adoption and Enablement

Technology adoption fails when it stops at training. My approach pairs role-based capability building with real workflow application: pilot cohorts apply AI to an actual piece of their job, not a generic exercise, so adoption is measured in changed behaviour and business outcomes, not attendance.

Advisory Principle

East-West AI Translation

Western AI governance frameworks and Chinese AI ecosystem developments are moving at different speeds with different assumptions. Bridging them requires native fluency in both the language and the regulatory logic - not just translation of words.

中文视角

跨界赋能者 | 让战略落地 | 让智能生效

为企业与初创公司，我提供可执行的转型路线图，从AI云战略到智能运营，并构建具备成本优势的韧性供应链。为教育机构，我通过将AI深度融入课程与评估，设计面向未来的教学模式。

双语AI业务融合：将AI与真实业务解决方案相结合，通过母语级中文专长进行AI模型训练、翻译质量保证和文化语境整合，化复杂为清晰。

EU AI Act guides

EU AI Act · Article 50

EU AI Act Article 50 in Ireland: What Applies From 2 August 2026

Chatbot disclosure and AI-content marking apply from 2 August 2026. What the Digital Omnibus deferred, what it did not, and who enforces it in Ireland.
EU AI Act · AI Literacy

EU AI Act Article 4: AI Literacy Training With Inspection-Ready Evidence

The AI literacy obligation has been in force since February 2025. What proportionate, role-based training means, and why records are the practical way to evidence it.

Featured writing

AI Governance · East-West

What the Gaokao AI Shutdown Taught the World About AI Governance

Five major Chinese AI platforms coordinated simultaneous feature shutdowns during China's national exam. What this means for AI governance globally.
Cloud Architecture · Media

Tech Meets Storytelling: How Your Videos Reach the World

Behind the Upload button is an orchestra of cloud services storing, moderating, processing and delivering video worldwide. What happens to your content, and what you can see about your audience.
Career · Strategy

Debugging Myself - Not Broken, Just Inefficient

On making the transition from enterprise execution to strategic advisory, and what it requires you to rebuild about how you operate.

Contact Me

Enquiries about advisory work, AI education programmes, speaking, or collaboration.

A small confession: I am not quite ready to hand my calendar over to an AI just yet. Every message here comes directly to me, read and responded to by a real human. Some things are still better that way.

0/1000