Freelance consultant to a London digital agency, delivering two production AI systems across a 12-month engagement. PROJECT 1 — Multi-Agent AI Automation System Designed and architected a multi-agent system handling WordPress content management, customer support, and marketing campaigns across multiple client accounts. Three domain-specific agents with confidence scoring, human escalation logic, and preview-before-publish controls. FastAPI + Redis Streams event orchestration, Pydantic state machines, multi-channel integration (web, email, WhatsApp). AI governance aligned with ISO/IEC 42001 and NIST AI RMF. Observability via Langfuse and Logfire. Results: 45% faster content delivery, 58% faster campaign turnaround, 41% automated support resolution with human quality gates. PROJECT 2 — B2B Sales Intelligence Platform (TradeReach) Led product strategy, UX architecture, AI system design, and design system end-to-end. Reframed the client brief from "work the list faster" to "what makes a call worth making?" — then shipped the platform in four months. Two-agent architecture: Scout Agent enriching leads across nine public signals with custom review intent classification (Gap Signal, Strong Reputation, Mixed Signal, Thin Data); Briefing Agent generating structured 30-second call briefs — priority reason, named pitch angle, talking track in the prospect's own customers' words. Autonomy dial for rep control, like/dislike feedback loop, editable scoring criteria, one-button recompute. Design system rebuilt from the token layer up: Figma tokens synced to Angular as CSS custom properties, WCAG AA semantic paired tokens, design.md governing AI-assisted code generation. All LLM calls traced in Langfuse: prompt version, token cost, latency, helpfulness and conversion tracked as independent signals. Results: 10× faster call preparation (5–10 min manual research → under 60 seconds), brief reframe to shipped product in 4 months (Angular, Supabase, LLM integration), agency began exploring the platform as an offering to other digital agencies before engagement closed.

Head of UX for a B2G/B2B assessment SaaS platform deployed in 70+ countries, across four continents. Built and led the design function over 10 years — from founding UX hire to strategic capability — supporting $1M to $20M revenue growth. TEAM & DESIGN OPERATIONS Built UX team from 0 to 5 designers, maintaining a lean 1:20 designer-to-engineer ratio across 100+ engineers. Established DesignOps frameworks: 4-phase product lifecycle embedding UX end-to-end, 75% QA autonomy through framework design, 90% team decision independence via progressive autonomy coaching. Led cross-functional delivery for 15-person teams; acted as interim Product Owner integrating UX into technical architecture and strategic decisions. AI-ASSISTED PRODUCT REDESIGN (2023–2024) Led redesign of the core authoring experience integrating AI-assisted workflows and familiar interaction patterns (Google Drive-like navigation). International validation: 400+ participants across EU, Japan, and US markets, 30+ in-depth interviews, mixed methods. Applied Triple Diamond methodology with IDEO DVF framework for systematic trade-off analysis. Built "Preview Alpha" prototype enabling simultaneous marketing demos and B2G hypothesis testing before committing engineering resources. Results: +16 SUS (68→84) validated in MEXT pilot (Japan Ministry of Education), 50% reduction in content creation time, 3× user adoption increase (A/B testing, n=400), strategic pivot validated saving 6+ months potential waste. DESIGN SYSTEM & ACCESSIBILITY Governed creation and evolution of 120+ component design system (Figma) across 6 SaaS products. Tokenized architecture: Minor Third 1.20 typographic scale, 8px vertical rhythm, systematic colour palettes. WCAG 2.2 AA compliance throughout. Four-phase accessibility validation: blind and low-vision users with assistive technologies, K-12 students across diverse learning contexts, national deployment testing (Norway UDIR, France DEPP), OECD PISA international coordination. Results: EdTech Breakthrough Awards 2022 finalist (top 10 / 2000+ solutions), 60% faster new product launches across 6 products, platform selected for OECD PISA international student assessment. PHYGITAL SERVICE DESIGN — FRANCE MINISTRY OF EDUCATION (2018–2019) Designed end-to-end national data collection service across physical and digital layers: OCR capture forms co-designed with psychometricians (zone sizing, contrast, QR placement per page), smartphone-based encoding pipeline replacing 3–4 hours of manual data entry, analytics portal with national benchmark overlays, and parallel monitoring interface spanning classroom teacher to regional authorities across France's administrative hierarchy. Results: 5× data collection time reduction (3–4h → 35–50 min, prototype-validated), 85%+ form responses auto-validated, SUS 81, deployed across thousands of classrooms nationwide.

Web development and semantic web technologies for public sector knowledge management at a national research center in Luxembourg. Designed and built interfaces connecting curriculum frameworks, educational content, and assessment outcomes using semantic knowledge graphs (SPARQL, RDF, RDFS, ontologies). Translated business and functional requirements into technical solutions through close collaboration with government stakeholders and engineering teams. Early foundation in structured data modelling, knowledge architecture, and B2G delivery — directly informing later information architecture and AI system design work.

Enterprise Design Thinking Co-Creator

IBM

2020

Credential ID: 859ff9f5-5cfa-4618-bac2-450d8c4d494d

Enterprise Design Thinking - Team Essentials for AI

IBM

Credential ID: bbe587ac-a519-4b06-bc57-bdd931f1cda7

Enterprise Design Thinking Practitioner

IBM

2020

Credential ID: 50a2a960-7fa2-485c-94c4-9ef6a0146d6e

Amplitude Foundations - Analytics

Amplitude

2025

Credential ID: 3651b707-df16-44f1-b894-81855710c70f