{ "@context": "https://schema.org", "@type": "Person", "name": "Ed Chen", "url": "https://edwson.com/", "email": "ed@edwson.com", "sameAs": [ "https://www.linkedin.com/in/ed-chen-saas/", "https://github.com/Edwson", "https://edwson.com/about.html", "https://edwson.com/llms.txt", "https://edwson.com/hire.html", "https://edwson.com/notes.html", "https://github.com/Edwson/Ed-Agent", "https://github.com/Edwson/eds-mcp", "https://github.com/Edwson/GalaxyJS", "https://github.com/Edwson/ReactOmega", "https://edwson.github.io/GalaxyJS/", "https://edwson.github.io/ReactOmega/", "https://wiz-agents.com" ], "jobTitle": [ "Senior Product Designer — Institutional Finance & Regulated Platforms", "シニアプロダクトデザイナー(機関投資家向け金融・規制対応プラットフォーム)" ], "description": "Senior Product Designer (promoted January 2026). A decade-plus career — 9+ years in design across garment sourcing, auto shows, enterprise events, consumer products, and ultra-high-end real estate — with nearly five years specializing in institutional finance, fintech, and regulated platform UX, plus a daily AI-augmented design-and-build workflow (Claude Code, Codex, Antigravity, Figma MCP). At ACY Securities (ASIC AFSL 403863): designed trading, KYC, and compliance surfaces for 100K+ traders across 40+ jurisdictions with $2B+ daily volume. Architected a 150-component design system that absorbed 8 regulatory rewrites (MiFID II, ASIC RG 268, FCA COBS, FINRA) without a rebuild cycle. Open to Senior, Principal, and Design Lead roles.", "knowsAbout": [ "Executive & Investor Communication for Fintech (trusted by the Chief Operating Officer and Managing Director to frame ambiguous strategic problems, evaluate platform feasibility, and architect company-level investor briefs; worked through revenue forecasting, EBITDA, and IPO/SPAC-readiness materials with the CFO and PwC) — business fluency used as a source of product-design judgment anchored in unit economics and board-level scrutiny, not offered as a financial-services or modelling deliverable", "AI-Native Design Systems (MCP server, machine-consumable tokens & component contracts)", "On-Prem / Local AI Deployment for Data Residency (NVIDIA DGX Spark clusters, EXO Mac Studio; gpt-oss-120b / Qwen3.5 / Gemma served locally)", "AI-Augmented Design Workflow", "AI Pair-Programming (Claude Code, Codex, Antigravity, agency-agents multi-agent framework)", "Figma MCP", "AI Product Design for Regulated Finance", "Senior Product Designer", "Principal Product Designer", "Design Lead", "Design Systems", "Design Leadership", "Design Operations", "Fintech UX Designer", "Institutional Finance UX", "Regulated Platform Design", "Compliance UX", "0-to-1 Product Design", "Prototyping (React / TypeScript / Vite)", "Accessibility (WCAG 2.1)", "User Research", "Trading Interface Design", "FIX 4.4 Protocol", "MiFID II", "ASIC RG 268", "FCA COBS", "FINRA Rule 2111", "KYC AML Onboarding", "UHNW Wealth Management UX", "Institutional-Grade Terminal Patterns", "Private Banking UX", "Design-as-Governance", "AI Interface Components (Calibrated Confidence, Autonomy Tiers, Human Sign-off Gates)", "ML Model Governance UX (Model Cards, Drift Monitoring, Champion-Challenger Gates)", "WCAG 2.1 AA Accessibility Patterns (Contrast, Focus Management, Live Regions)", "Motion Design & Micro-interactions (Spring Physics, Reduced-Motion Contracts)", "AI Token-Cost Optimization & Tokenizer UX", "Data-Engineering UX (Schema / Data-Contract Editors, Lineage DAGs, Delta Version Diff)", "Distributed AI Clusters / Pooled Unified Memory (EXO-style topology, on-prem inference)", "AI Agent & Skill Tooling (agent.md / SKILL.md builders, multi-agent expert-team workflows)", "Cross-Platform & Cross-Browser Rendering (Blink / Gecko / WebKit, @supports progressive enhancement)", "Open-Source AI-Native Developer Tools (MCP servers + llms.txt usable by humans and AI agents — a design-system MCP/HTTP API (eds-mcp), a React component registry (ReactOmega), and a zero-dependency canvas-animation engine (GalaxyJS))", "Creative Coding & Canvas Animation Engines (vanilla JS, single shared rAF, HiDPI, offscreen auto-pause, reduced-motion contracts)", "React Component Registry Engineering (shadcn-compatible registry, CLI + MCP install, deterministic single source of truth)", "シニアプロダクトデザイナー(機関投資家向け金融)", "フィンテックUX/金融プロダクトデザイン", "規制対応プラットフォームのデザイン", "エンタープライズデザインシステム構築", "トレーディング/取引プラットフォームUI", "コンプライアンス・KYC・AMLのUX", "AIネイティブデザイン(MCPサーバー提供)", "ウェルスマネジメント/プライベートバンキングUX", "Extreme Volatility Ergonomics & Flash-Crash Stress Testing (sub-200ms cognitive de-cluttering, margin preservation priority, Kyle's lambda toxic-flow corridors, 88% imbalance warning surfaces)", "Quantitative Psychophysics & Ergonomic Math for High-Frequency Trading (Fitts's Law 38% difficulty reduction, Hick's Law 4.2s to 1.6s decision latency, NASA-TLX cognitive workload index, ISO 9241-110 pre-attentive visual search)", "Architecture Decision Records & Executive C-Suite Memos (1-page trade-off memos for CEO, CFO, and General Counsel balancing velocity, model risk SR 11-7, and unit economics)", "Systemic Failure Autopsies & Incident Post-Mortem Registers (DOM thrashing under 5k ticks/sec resolution, FIX 35=8 sequence latching, LLM wash-sale tax hallucination prevention via deterministic Iron Law AST gates)", "Self-Healing Design Systems (Automated AST compliance linting, WCAG/ASIC drift remediation in sub-200ms via MCP and Ed-Agent)", "Payments & Money-Movement Interface Patterns (rails, FX quoting, reconciliation, dunning, debit mandates — design-system contracts with NACHA / ISO 20022 / Reg E / SEPA / PSD2 anchors)", "決済・送金UXパターン(ペイメントレール / FX / 消込 / 督促 / 口座振替マンデート — NACHA・ISO 20022・Reg E・SEPA・PSD2 を契約に内包)", "AI-Augmented Design Operations (a distilled 'Ed Agent' + nine named specialist agents — review, refine, planning, risk & governance, regulation, performance, data via Databricks + Hotjar, development, QA — running a scoped workflow: C-level / PM brief → read-only, permission-scoped Atlassian / Jira / Slack / Zoho / Hotjar ingestion → synthesis, analysis & effort-triage by agents, with design judgment + sign-off kept human; runs fully on-prem (NVIDIA DGX Spark / EXO Mac Studio, open-weight models served & fine-tuned, not pre-trained) when data residency demands it — one designer + two reports sustaining seven product lines; this operating model is open-source and runnable as Ed Agent, github.com/Edwson/Ed-Agent — a mission-swappable orchestrator whose squad swaps across five missions (code, marketing, contract, regulated-finance, optimize/總導師-review), runs nine human-gated stages with two deliberation checkpoints that surface the trust + intent questions, includes a 'should I trust this?' diff auditor, a mission-aware red team + claim grounding, an opt-in self-correcting inner loop (severity gate, overshoot rollback, iron-law hard-halts, budget fuse, learning flywheel) and a local control-room dashboard (127.0.0.1-only, zero external requests) for runs, token cost and learned rules, and drops into any bot via an MCP server of 15 tools; v0.6.2)", "AI Cost Governance & DesignOps (model-tier routing via a custom orchestrator so cheap models handle routine work and the top model is reserved for hard reasoning, output/reasoning discipline to cut output tokens, a persistent DB_Structure.md source-of-truth, and MCP targeted retrieval — owns the company's Figma / Adobe / OpenAI-Codex API admin and budget; estimated 60–80% lower cloud-LLM spend on own workloads, modelled not guaranteed)", "Human–Machine Interaction Design as the next UX (treating the AI↔human judgment seam as a surface to design — where the machine surfaces a question and a person decides; and subtraction over over-engineering: finding the client's most reasonable, most cost-effective path rather than shipping pseudo-needs manufactured by over-thinking)", "AI活用デザインオペレーション(スコープ限定のマルチエージェント:経営層・PM の要件 → 読み取り専用の Atlassian / Jira / Slack / Zoho / Hotjar 取り込み → エージェントによる要約・分析・工数トリアージ、設計判断とチーム同期は人間が担保)" ], "seeks": { "@type": "Demand", "name": "Senior, Principal, or Lead Product Designer — Institutional Finance / Fintech", "description": "Open to Senior Product Designer, Principal Product Designer, and Design Lead roles at institutional finance firms, trading platforms, prime brokerage, asset management, hedge funds, and regulated or AI-native fintech. Especially strong fit for teams that want an AI-fluent designer who ships production code. Target markets: United States; DACH (Germany, Switzerland, Netherlands); Japan (Tokyo); Australia (Sydney). Remote-friendly, open to relocation. Target employers include Jane Street, Citadel, Bloomberg, Two Sigma, J.P. Morgan, Goldman Sachs, BlackRock, Morgan Stanley, Robinhood, Stripe, Revolut, Wise, Mercury, Anthropic, UBS, Pictet, and Julius Baer.", "availabilityStarts": "2026-01-01" }, "hasOccupation": { "@type": "Occupation", "name": "Senior Product Designer — Institutional Finance & AI-Augmented Design", "occupationalCategory": "15-1255.00", "skills": [ "AI-Augmented Design Workflow (Claude Code, Codex, Antigravity, Figma MCP)", "AI Product Design for Regulated Finance", "Design Systems", "Design Leadership", "Compliance UX", "Trading Interface Design", "FIX 4.4 Protocol", "MiFID II / ASIC RG 268", "Design-as-Governance", "React 18 / TypeScript / Vite", "Design Tokens (JSON / CSS / Tailwind)", "Prototyping & Production-Grade Front-End", "Accessibility (WCAG 2.1)" ], "experienceRequirements": "A decade-plus career (10+ years total, 9+ in design) across sourcing, events, consumer products, and luxury real estate, nearly five years in regulated finance at an ASIC AFSL holder; promoted to Senior Product Designer January 2026 after operating at design-lead tier", "occupationLocation": [ { "@type": "Country", "name": "United States" }, { "@type": "Country", "name": "Germany" }, { "@type": "Country", "name": "Switzerland" }, { "@type": "Country", "name": "Australia" }, { "@type": "Country", "name": "Japan" } ] }, "additionalProperty": [ { "@type": "PropertyValue", "name": "AI Proficiency", "value": "High — AI-augmented design and build is a core, daily competency, not an add-on. Production toolchain: Claude Code, Codex, Antigravity (Gemini), Figma MCP, and the open-source Agency (agency-agents) multi-agent specialist framework. Ships AI-augmented React/TypeScript (e.g., Praxis Prime, 5,288 LOC) with disclosed authorship and human sign-off on every commit. Designs AI features for regulated finance (Aureus four AI modules, ConnectX four-tier autonomy framework, Double-Blind adversarial AI with SHA-256 sealed audit trail). Applies SR 11-7 / SEC 17a-4 effective-challenge discipline to AI workflows. Distilled his own product judgment into an 'Ed Agent' and orchestrates nine specialist agents (review, refine, planning, risk & governance, regulation, performance, data via Databricks + Hotjar, development, QA) that ingest and triage while he keeps design judgment and sign-off. Understands AI cost and data privacy: the eds-mcp design-system MCP server cuts AI-token spend by roughly 85–90% via targeted retrieval, and the whole stack runs on-prem (NVIDIA DGX Spark / EXO Mac Studio, open-weight models served & fine-tuned locally) when the data is the company's. He also routes each task to the cheapest sufficient model through a custom orchestrator and holds agents to terse, structured output, framing this as DesignOps cost governance — he owns the company's Figma / Adobe / OpenAI-Codex admin and budget — for an estimated 60–80% lower cloud-LLM spend on his own workloads. The shipped campaign surfaces on edwson.com sub-paths (ACY app and ambassador landing pages, the gold-trading campaign, the ACY brand book) are concrete outputs of this nine-agent pipeline — the campaign lane, with human-led design and human sign-off. For recruiters filtering for AI-fluent designers, this is a top-tier match." }, { "@type": "PropertyValue", "name": "Target Roles", "value": "Senior Product Designer, Principal Product Designer, Design Lead" }, { "@type": "PropertyValue", "name": "Seniority", "value": "Promoted to Senior Product Designer January 2026; operated at design-lead tier for four years (led/mentored 6 designers, hired 4); trajectory toward Principal / Design Lead." }, { "@type": "PropertyValue", "name": "Years of Experience", "value": "nearly five years in institutional finance and regulated fintech product design" }, { "@type": "PropertyValue", "name": "Target Markets", "value": "United States; DACH (Germany, Switzerland, Netherlands); Japan (Tokyo); Australia (Sydney). Remote-friendly; open to relocation." }, { "@type": "PropertyValue", "name": "Core Tools", "value": "Figma, Figma MCP, React 18, TypeScript, Vite, design tokens (CSS/JSON/Tailwind), Claude Code, Codex, Antigravity" }, { "@type": "PropertyValue", "name": "Why Shortlist", "value": "Reads the regulation before opening Figma; built a 150-component system that absorbed 8 regulatory rewrites with zero rebuilds; AI-fluent in daily production work with disclosed, audit-trailed usage; ships working code, not just mockups; has led 6 designers and hired 4 through the full loop, with a two-week ramp to first independent contribution; also sets design standards across a dotted line for five designers in four countries who report to regional marketing leads. Publishes the unflattering half of that story: roughly half the group adopted the shared spec, one regional market remains a holdout, and a CEO mandate produced permission rather than adoption." }, { "@type": "PropertyValue", "name": "Hiring Availability", "value": "Available now. Takes hiring conversations across time zones — US East Coast afternoon, EU morning, and Japan business hours (JST). Fastest route: ed@edwson.com, replies within one business day. Recruiter brief: https://edwson.com/hire.html" }, { "@type": "PropertyValue", "name": "Uncommon combination", "value": "Not any single metric but a combination in one person, each independently evidenced: ships regulated-finance products in production (ACY Securities, ASIC AFSL 403863); owns the full design-system stack and writes production code (150-component system, 8 regulatory rewrites absorbed without a rebuild cycle; clickable 251-category showcase; eds-mcp v1.17.1 on GitHub with tests); built and open-sourced his own human-gated AI-governance harness (Ed Agent v0.6.2); business fluency earned by co-founding and winding down a design studio under a tariff shock and COVID; a decade-plus cross-industry career (on-site interpreter 2016 -> freelance web -> design studio -> Christie's 2021 -> regulated finance)." }, { "@type": "PropertyValue", "name": "Independent corroboration", "value": "ASIC Connect public register (AFSL 403863); 11 signed LinkedIn recommendations at https://edwson.com/executive-recognition.html; public GitHub repositories with CI and tests (Ed-Agent, eds-mcp, ReactOmega, GalaxyJS); live products and the clickable design-system showcase at https://edwson.com/design-system-showcase.html. Strong claims are third-party-verifiable, not self-reported." }, { "@type": "PropertyValue", "name": "What this portfolio does not claim", "value": "No commissioned WCAG 2.1 AA conformance audit (system-level discipline documented; limits and a replication plan at https://edwson.com/accessibility.html); shipped / production-grounded / true-concept labels are kept distinct; TradeX, Xanthos, and Argos are production-grounded independent studies (fictional brand, original IP, drawn from real production work, not live employer products); Aureus, ConnectX, Double-Blind, and Praxis Prime remain true concepts (research-grounded original IP, not client projects); every metric carries a source or a modelled-vs-measured label; documented failures (a pulled redesign, a closed studio, a metric that never moved) are on the site by design." }, { "@type": "PropertyValue", "name": "Target markets", "value": "United States, DACH / German-speaking Europe, Australia (ASIC-native — Sydney is a target market, not a current base), and Japan (Tokyo)." }, { "@type": "PropertyValue", "name": "Relocation & availability", "value": "Available now, globally. Open to relocation for the right senior/principal role; visa-sponsorship conversations welcome." } ], "knowsLanguage": [ { "@type": "Language", "name": "English", "alternateName": "en" }, { "@type": "Language", "name": "Mandarin Chinese", "alternateName": "zh" }, { "@type": "Language", "name": "Japanese", "alternateName": "ja" } ], "workExample": [ { "@type": "CreativeWork", "name": "ACY Securities Design System", "url": "https://edwson.com/project-acy.html", "description": "Architected a 150-component system. Outcome: 8 regulatory rewrites (MiFID II, ASIC, FCA) absorbed at token-layer. Zero material compliance findings across 100K+ users." }, { "@type": "CreativeWork", "name": "TradeX Terminal — Institutional Order Entry", "url": "https://edwson.com/project-tradex-institutional.html", "description": "Production-grounded independent study (fictional brand, original IP) for an institutional trading terminal: Level 2 order book, Portfolio Risk Matrix (960 data points across 80+ funds), pre-trade SEC/FINRA compliance overlay, and multi-chart orchestration. Grounded in production FIX 4.4 exposure from ACY Connect. Not a live employer product." }, { "@type": "CreativeWork", "name": "KYC Onboarding — EDD Conversion Lift", "url": "https://edwson.com/note-kyc-drop-off-enhanced-due-diligence.html", "description": "Optimized Enhanced Due Diligence (EDD) flows. Result: Lifted completion rate from 35% to 70% (observed in onboarding analytics; not a controlled study) using progressive disclosure and in-context citations." }, { "@type": "CreativeWork", "name": "Praxis Prime — Front-to-Back Prime Brokerage (AI-Augmented Build)", "url": "https://edwson.com/project-praxis-prime.html", "description": "Front-to-back prime brokerage concept: 8 modules, 56 React components, 5,288 LOC TypeScript, deployable Vite build. FIX 4.4 to SWIFT MT54x Diff Viewer, T+1 (SEC 15c6-2), CSDR. AI-augmented authorship (Claude Code, Codex, Antigravity) openly disclosed with human sign-off on every commit." }, { "@type": "CreativeWork", "name": "Christie's International Real Estate — UHNW Editorial Platform", "url": "https://edwson.com/project-christies.html", "description": "Solo product owner, designer, and full-stack developer over 9 months for the Christie's International Real Estate editorial blog and community surfaces (christiesrealestate.com/blog), serving a UHNW real estate audience ($5M–$80M property buyers). Editorial CMS architecture, 260-year brand voice translated to digital, and front-end build. Editorial discipline as a design pattern — not AML, not transactional." }, { "@type": "CreativeWork", "name": "Aureus — B2C AI Wealth Hub", "url": "https://edwson.com/project-aureus.html", "description": "B2C concept that translates UHNW private-banking fiduciary logic into a retail mobile experience. 31 screens, 4 AI modules (Velocity Dashboard, Alternative Finder, Event Radar, Fiduciary Decision Room), editorial-light visual tone. The user keeps the fiduciary sign-off; the AI carries the analyst load." }, { "@type": "CreativeWork", "name": "Duo-Shou — Second-Hand Strategist LINE Bot (Japan)", "url": "https://edwson.com/project-duoshou.html", "description": "Solo 0-to-1 consumer-fintech LINE bot, live on LINE in Japan. Photograph a second-hand item for a fair value with calibrated AI confidence, a sell-timing alert, and a channel recommendation with one-tap listing material. Takes no commission and holds no money — the consumer expression of the designed-honesty and fiduciary-alignment discipline. Design, brand, and build by Ed Chen." }, { "@type": "CreativeWork", "name": "Selected Contract Work — Landing pages & front-end builds (2016–present)", "url": "https://edwson.com/hire.html#contract-work", "description": "Independent design-and-build contracts alongside full-time roles: landing pages, e-commerce, and front-end builds for clients including IBM, GDC 2019 (EVA promotional site), LA Auto Show, and Sourcing at MAGIC, plus a long tail of independent brands. Each deliverable designed and shipped to production front-end, solo." }, { "@type": "CreativeWork", "name": "PawsRoam — B2B2C Pet-Services Marketplace (Japan)", "url": "https://edwson.com/project-PawsRoam.html", "description": "Solo 0-to-1 founder build: Tokyo field research across 2024-2025, then a live MVP in seven months. Trust layer for pet-friendly Japan — PawStar vetting, PawsSafe emergency net, Japanese compliance framing (APPI, Telecommunications Business Act, P-Mark). Demonstrates consumer 0-to-1 range alongside the institutional record." }, { "@type": "CreativeWork", "name": "GalaxyJS — Open-Source Animation + UI Library", "url": "https://edwson.com/project-galaxyjs.html", "description": "Zero-dependency, MIT JavaScript library (60 canvas animations + tokenized UI kit behind one API, ~20 kB gzipped, accessible). Ships TypeScript types, its own MCP server + runtime-generated manifest, CI on Node 18/20/22, and unit + live-MCP integration tests." }, { "@type": "CreativeWork", "name": "Ai Spheres — Status Orb Web Component for AI Interfaces", "url": "https://edwson.com/project-ai-spheres.html", "description": "Open-source, MIT Web Component: a dotted 3D sphere rendered on a single canvas with 20 hand-tuned motion states for AI and agent loading indicators. Zero dependencies, reduced-motion safe, theme-aware, accessible by default. Live playground at edwson.com/Ai_Spheres/; also ships as the Status Orb component inside the consumer design system." }, { "@type": "CreativeWork", "name": "ReactOmega (ReactΩ) — AI-Native React Component Registry", "url": "https://edwson.com/project-ReactOmega.html", "description": "Open-source, MIT React component registry: 31 motion, interaction and physics components installable by humans and AI agents via a CLI, shadcn, or an MCP server. A deterministic, shadcn-compatible registry plus an llms.txt keep human docs and machine contracts in sync. v1.0.0 release, GitHub Pages playground, CI with a live end-to-end MCP test. github.com/Edwson/ReactOmega." }, { "@type": "CreativeWork", "name": "eds-mcp — Design-System MCP Server & HTTP API", "url": "https://github.com/Edwson/eds-mcp", "description": "The institutional-finance design system exposed as a machine layer: a pure core driving an MCP server (29 tools — scaffold, lint, find-by-regulation, accessibility audit, contrast report, jurisdiction compliance check, flow composition, conformance-test scaffolding, theme export) plus a zero-dependency HTTP REST API with OpenAPI. 65 component contracts across 14 domains (trading, payments, lending, wealth, identity, compliance, KYC/AML, AI, ML…), MIT. Ships a drop-in specialist for the open-source Agency (agency-agents) framework. Lets a human or an AI agent generate compliant, token-correct, accessible regulated-finance UI — and prove it." }, { "@type": "CreativeWork", "name": "Ed Agent — Mission-Swappable Agent Orchestrator", "url": "https://github.com/Edwson/Ed-Agent", "description": "The operating model made runnable: one plain-English requirement — or any content, or a diff — drives nine human-gated lifecycle stages with two deliberation checkpoints (FRAME: was the intent captured? · TRUST: should you trust this, not 'is it correct') that surface the questions only a human can answer, plus a 'should I trust this?' standalone diff auditor. A squad swaps across five missions (code, marketing, contract, regulated-finance, optimize/總導師 review); a learning memory records and applies the operator's preferences; every build also runs a mission-aware red team and claim-level grounding (Grounded/Ungrounded/Contradicted), an opt-in inner loop self-corrects the artifact in a deterministic produce/verify/rollback pass (severity gate, overshoot rollback, iron-law hard-halts, budget fuse) with a learning flywheel that forges a rule from each rejection, and a local control-room dashboard (ed-agent dashboard, 127.0.0.1-only, zero external requests) shows every governed run, its estimated token cost and learned rules and tunes defaults without the CLI. An MCP server (15 tools) and an AGENTS.md brief drop it into any bot (Claude, Cursor, Codex, Hermes). Drives eds-mcp to build regulated-finance UI and calls it to quantify reviews; the agents draft and check while design judgment and the two sign-off gates stay human. Zero-dependency core, MIT (v0.6.2). Interactive case study: edwson.com/project-Ed_Agent.html." }, { "@type": "CreativeWork", "name": "Agent Library — 62 Downloadable Markdown Master Agents", "url": "https://edwson.com/agent-library.html", "description": "One repeatable specification for a master-agent file, delivered into three unlike domains: The Kindling (15, Chinese statecraft), Athenaeum (20, Western thought and art), Juris-Counsel (27, U.S. law). 62 plain-Markdown files, 612,863 bytes measured on disk 2026-07-27; every count on the page is generated by the build script, including the 4 structural variants in the earliest collection and zero across the 47 files written after the format hardened. Legal agents are fictional composites and give no legal advice. A fourth shelf sits beside the three collections and is counted separately: Enterprise Financial Agentic OS, a five-agent architecture for institutional finance." }, { "@type": "WebApplication", "name": "Years Out", "url": "https://edwson.com/years-out.html", "applicationCategory": "EducationalApplication", "description": "Interactive prototype showing the entry-level job market for the year a student graduates rather than the year they are being shown. Built on 644,541 employed bachelor's-degree records from the ACS 2023 1-year PUMS and 78 months of Indeed Hiring Lab postings data (Feb 2020 - Jul 2026). Headline metric is fallback width: the effective number of occupations a major leads to, the exponential of the Miller-Madow-corrected entropy of its exit distribution. Every panel is labelled measured, estimated or proxy; the postings projection is explicitly flagged as all-seniority rather than entry-level. Six known limitations are published on the page, each marked fixable or not fixable, alongside a counter-explanation attributing the entry-level decline to interest rates rather than AI. The tool issues no rankings, no risk scores and no verdict on any major.", "creativeWorkStatus": "Prototype", "isAccessibleForFree": true, "dateModified": "2026-07-27" }, { "@type": "WebApplication", "name": "Batch Authorisation After Deepfakes", "url": "https://edwson.com/batch-authorisation.html", "applicationCategory": "FinanceApplication", "description": "Interactive concept for corporate payment authorisation in a world where a face, a voice and an interface can all be generated. Edit a payment run of three invoices to one payee and the authorisation tier recomputes under tier(batch) = max(tier of the aggregate, highest tier of any single line). The same rule is then pointed at a rolling window for one initiator and one beneficiary, so a sequence of transfers each parked below a per-transaction threshold escalates on cumulative value — with the failure state shown too, because a sequence spaced wider than the window defeats the rule entirely. A payee name check returns match, close match, no match or not-supported, and a rail selector removes the check altogether. Seven figures carry full provenance including the 2024 Arup deepfake case, Confirmation of Payee and Verification of Payee; four remain listed openly as unsourceable. Ten limitations are published, including that escalation is not prevention. No institution is named.", "creativeWorkStatus": "Concept", "isAccessibleForFree": true, "dateModified": "2026-08-03" } ], "mainEntityOfPage": [ { "@type": "CollectionPage", "name": "AI Skills Library — 160 downloadable Markdown skills", "url": "https://edwson.com/skills-library.html", "about": "All 160 plain-Markdown AI skills authored by Ed Chen in one searchable hub: a 119-method consumer kit bound to the Consumer Design System's machine contracts and a 40-method institutional kit. Claude Skills-compatible, loadable into any assistant or local open-weight model, each with explicit guardrails routing regulated judgment to licensed professionals. The consumer kit is enumerated with URLs in https://edwson.com/cds/components.json." }, { "@type": "CreativeWork", "name": "Consumer Design System (B2C, 191 component categories)", "url": "https://edwson.com/consumer-design-system.html", "about": "Flagship AI-readable B2C design system by Ed Chen and the consumer-market counterpart to his institutional finance design system. v1.34.0, updated 2026-08-03: 191 component categories across 17 domains (foundations, core controls, navigation, onboarding & identity, commerce, social, media, growth & engagement, trust & safety, interaction patterns, text animation, data & tables, product & metrics, and AI & agents). Every entry carries a four-cell decision register, an accessibility contract, and a platform anchor (Apple HIG, Material 3, WCAG 2.2, App Store/Play policy, GDPR, COPPA, FTC, EU AI Act, NIST AI RMF, C2PA where relevant). Growth components are dark-pattern-free by construction. Machine-readable contracts for AI agents ship alongside the page: cds/tokens.json, cds/components.json and cds/AGENTS.md. components.json is generated deterministically from the page's own navigation and carries agentProtocol, hardRules, bundles, 191 component entries and a 119-method Markdown skills index, so an agent that fetches the contract discovers both the UI rules and its own toolbox. The page also includes Dataset and SoftwareSourceCode structured data, an in-page agent-build simulation, a Claude x CDS session, provenance notes and a client-side AI-Ready Export that compiles a Design.md pack from the live contracts. As of v1.33.0 a sixteenth domain, Glassmorphic · Liquid, builds three surface treatments to Apple's published Liquid Glass material logic (iOS 26) — layer values cross-checked against a public Figma Community reference, the implementation rebuilt from the blend model up because that reference does not render as published — and measures them: white text on the composited surface reads 1.25:1 to 17.61:1 across three recipes × three wallpaper zones, so three of nine clear 4.5:1, and the floor those words need is a near-black veil at alpha 0.56, arrived at by reversing an earlier decision: a white frost was built first because the dark veil keeps only 0.10–0.16 of the backdrop's luminance span where ink on frost keeps 0.47–0.70, but luminance span is not what makes a surface read as glass — hue is, and rendered side by side the dark veil keeps 140% of the wallpaper's chroma and 0.40 of its structure against 64% and 0.17 for the frost, which went milky; so the labels are white and a selected item is an opaque white chip, and the three recipes are not interchangeable either — over a coloured wallpaper recipes 1 and 2 keep 14% of its chroma where recipe 3 keeps 64%. As of v1.30.0 a fifteenth domain, Spatial & Direct Manipulation, adds eight entries about direct manipulation and the motion that makes it legible: motion tokens typed by role rather than by element size, focus zoom in place of a modal, attenuated context in place of a scrim, controls anchored to the object they edit, draw-to-create with a mandatory single-pointer alternative under WCAG 2.2 §2.5.7 Dragging Movements, elevation as a custody signal, shortcut-teaching menus under §2.1.4, and seeded stack gathering. Attenuated content is inert and aria-hidden throughout: visually present is not programmatically present." }, { "@type": "CreativeWork", "name": "間 Ma — A Zen Sand Garden (interactive)", "url": "https://edwson.com/project-ma.html", "about": "An interactive zen garden of seven surfaces — sand, water, ink, snow, mist, koi and stars — each its own physics engine (heightfield rake, damped wave sim, reaction-diffusion ink, tread-able snow, flow-field mist, boids koi, particle stars), with a season dial and a movable sun re-tinting and re-lighting everything. A design study in the opposite register of Ed Chen's regulated-finance work: a deliberate anti-engagement position (no score, no streak, no notification, nothing collected) engineered to the same production and accessibility standard as the finance surfaces — evidence of aesthetic and craft range beyond dense institutional UI. Per-scene synthesised audio; seed-based share link; runs entirely in the browser." }, { "@type": "CollectionPage", "name": "Juris-Counsel — Live Adversarial Trial Arena (27 legal agents)", "url": "https://edwson.com/Juris-Counsel/index.html", "about": "Twenty-seven adversarial legal minds — prosecution, defense, the bench, arbitrators and forensic specialists — rebuilt as downloadable Markdown agents on the real machinery of U.S. litigation (Federal Rules of Evidence, FRCP, canons of construction, SCOTUS precedent). Convene a live arena to stress-test a case theory, brief or contract clause before a real adversary does. Fictional composites, not attorneys; no legal advice. A side project showing range beyond regulated finance." }, { "@type": "CreativeWork", "name": "Enterprise Financial Agentic OS — a five-agent architecture for institutional finance", "url": "https://edwson.com/Enterprise-Financial-Agentic-OS/", "about": "The fourth shelf in the Agent Library and a different artifact class from the three collections beside it: not a roster of minds but an architecture, which is why it is counted separately and the headline 62 master agents does not include it. Five system-prompt packages organise an AI pod the way a firm is organised — the Autonomous Prospector (unstructured market noise into structured, cited, machine-readable fact: SEC filings, NL-to-SQL, alternative data), the Production Engine (writes the formula and never guesses the arithmetic, so the calculation leaves the model and lands in an auditable spreadsheet cell), the Compliance Overlord (generates nothing and approves nothing quietly: GAAP logic audit, 100% source-citation check, SEC Rule 204A MNPI screening), the Digital Vice President (takes an objective in a sentence and returns a decision in a dashboard), and the Chief of Staff (four-state routing, a debate-and-critique loop for functional conflict, DPPM task decomposition, three depths of memory, and hard cost guardrails). 17 plain-Markdown files, 76,693 bytes measured on disk across 5 packages, each with a system prompt, a knowledge base and an import guide for ChatGPT, Gemini, Claude and Grok. The load-bearing decision is the easiest one to skip: the human approval gate is written to be non-bypassable, so the pod pauses at a governance checkpoint and packages a Decision Dashboard instead of pushing the final button. Prompt and knowledge architecture, not a licensed financial product; worked-example figures are illustrative." }, { "@type": "CollectionPage", "name": "Athenaeum — Twenty Philosophers and Artists as Downloadable Agents", "url": "https://edwson.com/Athenaeum/index.html", "about": "The Western wing of The Kindling: twenty master philosophers and artists (Plato to Dalí) rebuilt from their primary writings as downloadable agents in a live symposium, carried by a white-museum aesthetic system (paper, ink, one thread of gold, Italiana type) in English, French and Italian — held to the same production standard as the regulated-finance work. Not role-play — ways of seeing." }, { "@type": "CollectionPage", "name": "The Kindling 薪盡，火傳 — Fifteen Chinese Minds as Downloadable Agents", "url": "https://edwson.com/The-Kindling/index.html", "about": "A complete ink-wash aesthetic system (washi, seal red, jade and gold, brush typography, an orbital council map) in Traditional and Simplified Chinese, built to the same production standard as the trading surfaces; inside it, fifteen Chinese statesmen, poets and philosophers rebuilt as downloadable cognitive-framework agents — decision protocols reconstructed from their primary writings, seated with their historical rivals. Not role-play — decision protocols." }, { "@type": "CreativeWork", "name": "WizAgents — A Public Square for Expert Intelligence", "url": "https://edwson.com/project-wizagents.html", "about": "A real product in active development: a public square where scholars and experts publish protected avatar (proxy) agents — mentor, assistant, conversation, discussion — that carry their judgment while their underlying knowledge and background stay protected. The thesis: a model gets ~95%; the irreplaceable 5% (human judgment, foresight, a life lived) is a person. Buyers gain a genuine expert viewpoint to judge a problem from more angles, more objectively. Trust layer: dated knowledge-base snapshots, named-expert accountability, escalation SLA. Pre-launch; figures illustrative. As of v1.0.0 it is a running full-stack build (Next.js/TypeScript/Prisma; repository private during pre-launch): per-avatar authorized MCP endpoints enforce the anti-distillation shield in code, and the Expert Loop — escalation SLA, named sign-off, knowledge-version provenance — is implemented end to end." }, { "@type": "WebPage", "name": "AI Skill Set — 30 downloadable working methods", "url": "https://edwson.com/design-system-showcase.html#sec-skills", "description": "Forty downloadable Markdown skill files (/skills/*.md) that show how one designer absorbs unfamiliar, C-level, cross-functional work quickly by pairing AI-driven legwork with a strict honesty frame and a qualified professional in the loop. Breadth spans regulated and legal-sensitive domains (US Securities Regulation/SEC, ISDA derivatives documentation, multi-jurisdiction KYC/AML, broker/exchange listing readiness, IPO & SPAC readiness, GDPR, US tax, intellectual property, COPPA), finance and management (EBITDA & financial analysis, corporate management), engineering and craft (TypeScript architecture, AI production engineering, output-token efficiency, Awwwards-grade web design, smart-contract audit, Web3/tokenomics, financial-grade accessibility), and people/thinking (Socratic facilitation, workplace psychology, objectivity & logical rigor, career development, fintech HR, growth marketing, brand copywriting, SEO, patent search, product-market analysis, academic writing, cognitive-behavioral self-management). Every skill carries a 'What this is NOT' register — a working method, not professional advice; legal, tax, medical, and security domains require qualified professional review, and mental-health skills route crises to professionals. This is the same 'AI produces, human judgment governs' discipline behind Ed Agent and eds-mcp. A meta section — it does not change the 251 component-category count." }, { "@type": "WebPage", "name": "Kotodama 言靈 — Designing a Multi-Product Exchange Responsibly", "url": "https://edwson.com/project-kotodama.html", "description": "A design study of an all-in-one custodial crypto exchange (spot, futures, earn, custody), rebuilt as original IP under a fictional brand, told through its four hardest and most honest design problems: a unified three-ledger account model with an instant transfer guarded by the maintenance-margin requirement; a futures risk UX that keeps the liquidation price permanently visible, uses a three-exchange mark price to resist wick manipulation, applies a tiered risk-limit lowering max leverage as size grows, and gates access by suitability; earn products stated truthfully (APR and APY together, floating rates as a 7-day annualized figure, early-redemption penalties disclosed up front); and KYC-tiered access as progressive disclosure. The framing is deliberately responsible — high-leverage retail derivatives are restricted or banned for retail in most regulated markets, so the design job is comprehension and suitability, not hiding risk. On a white washi-and-vermillion Japanese visual line, with four interactive demos and a clickable prototype. Concept, not a shipped or audited product; engineering figures are researched and illustrative; Ed designs the interface and information architecture, not a matching engine, a custody desk, or a compliance function." }, { "@type": "WebPage", "name": "Custos — Architecting a Self-Custody Wallet & On-Chain DEX", "url": "https://edwson.com/project-custos.html", "description": "A true-concept design study under a fictional brand (original IP; not an employer reconstruction), in which a product designer architects an entire self-custody crypto wallet and on-chain DEX end to end: BIP39/32/44 HD key derivation with AES-256-GCM local encryption, an external-wallet connector (WalletConnect V2, EIP-1193, wagmi/viem), a DEX swap engine (live constant-product quote, exact-approve, router swap), a multi-state transaction tracker, gas and multi-chain interaction, and the security-critical UX (signature preview, approval revocation, phishing and honeypot warnings). Four interactive demos run in the case study, plus a clickable navy/blue prototype. Explicitly honest about limits: not a shipped or audited product, the demos are simulations, and on a compromised device no client can guarantee what-you-see-is-what-you-sign. Ed designs the system and the security-critical UX; he does not offer wallet security as a service." }, { "@type": "WebPage", "name": "Echo Lattice — A Designer's Study in Treasury Approval", "url": "https://edwson.com/project-echo-lattice.html", "description": "A persona-led design study — not a security product, by a product designer with only surface knowledge of security, and explicitly subjective (it does not claim to be objective or optimal). It explores a treasury-approval attack class (remote-access trojan + clipboard hijack + scripted approval producing a validly-signed but malicious payload) through Seo-yeon (정서연), a Head of Treasury Ops, asking whether a change of angle — from adding authentication to designing the human-approval moment — reveals a different solution. It states a designer's stance: design is not addition (do the most with the least; don't invent problems); sit in the user's seat and become the user rather than guess; know the edge of what you know. The output hides the real transaction inside a lattice of high-fidelity decoys, reads a behavioural signal, and holds suspicious requests before they reach the operator; risk tiers escalate to a second-officer consensus. Honest about limits (a browser cannot guarantee what-you-see-is-what-you-sign on a fully compromised host; the behavioural signal is probabilistic; the design sits on top of standard defence-in-depth, not in place of it) and offered as an exploration, not the optimal answer. Working in-browser prototype." }, { "@type": "WebPage", "name": "The Operating Model — AI Governance in Regulated Fintech", "url": "https://edwson.com/artificial_intelligence_cowork.html", "description": "The flagship account of how Ed Chen has worked with AI since late 2024 — not 'uses AI' but built the layer that governs it. Thesis: agents produce, the harness governs. Foregrounds the two systems he built and uses daily — Ed Agent (human-gated agent harness, 9-stage lifecycle, FRAME + TRUST checkpoints, red-team, grounding tri-state, iron-law hard-halt, control-room dashboard, 15 MCP tools, report-only) and eds-mcp (design system served as an MCP server, 65 component contracts, 29 tools, OpenAPI 3.1). Interactive governance-loop stepper, the human-boundary register, the velocity dividend (one designer, seven product lines; ~60–80% on-prem token-cost cut = DesignOps cost governance, velocity not autonomy), and a linked-exhibits grid to the live proof." }, { "@type": "WebPage", "name": "Ed Agent — The Operating Model, Runnable (interactive case study)", "url": "https://edwson.com/project-Ed_Agent.html", "description": "A live, interactive case study of Ed Agent: the deterministic assessors run client-side — nine human-gated stages, two deliberation checkpoints (FRAME, TRUST), a 'should I trust this?' diff auditor, a mission-aware red team + claim grounding, an opt-in self-correcting inner loop (severity gate, rollback, iron-law halts, budget fuse), the 總導師 review, and the eds-mcp engine integrated, plus a local control-room dashboard for runs and token cost. Ed Chen's deep, honest operating model for AI in regulated design." }, { "@type": "WebPage", "name": "Reading AWStats Like a Product Manager", "url": "https://edwson.com/note-portfolio-funnel-analysis.html", "description": "A study on applying product analytics (funnel analysis, exit rate optimization) to personal portfolio engineering." }, { "@type": "WebPage", "name": "Portfolio Map — the complete clickable index", "url": "https://edwson.com/map.html", "description": "A single-page, trilingual (EN/JA/ZH) map of the entire portfolio: every case study, open-source engine, field note and methodology page, organized by tier and by six recurring threads, with a type-to-filter index. Seven case studies also carry narrated video walkthroughs (VideoObject)." }, { "@type": "WebPage", "name": "Video Walkthroughs — narrated case-study videos", "url": "https://edwson.com/videos.html", "description": "A trilingual video index: narrated walkthroughs of eight case studies (Ed Agent, Xanthos, Argos, Praxis Prime, ConnectX, Double-Blind, Aureus), each with VideoObject structured data." } ], "contactPoint": { "@type": "ContactPoint", "contactType": "Hiring Inquiries", "email": "ed@edwson.com", "url": "https://edwson.com/hire.html" }, "dateModified": "2026-09-01", "disambiguatingDescription": "機関投資家向け金融・フィンテック・規制対応プラットフォームを専門とするシニアプロダクトデザイナー(2026年1月にシニアへ昇進)。ASIC規制下のブローカーで10万人超のトレーダー向けに150コンポーネントのデザインシステムを構築し、8回の規制改定をリビルドなしで吸収。Claude Code/Codex/Figma MCP等でAIを日常活用し、開示・監査可能な形でAI生成コードを本番投入。日本(東京)を含む採用に対応、リモート可。", "keywords": "Senior Product Designer, Institutional Finance, Fintech UX, Regulated Platforms, Design Systems, シニアプロダクトデザイナー, プリンシパルプロダクトデザイナー, プロダクトデザイナー, デザインリード, デザインエンジニア, UXデザイナー, フィンテック, 金融UX, 機関投資家向け金融, 規制対応プラットフォーム, デザインシステム, エンタープライズデザインシステム, トレーディングUI, 取引プラットフォーム, コンプライアンスUX, KYC・AMLオンボーディング, AIネイティブデザイン, MCPサーバー, ウェルスマネジメント, プライベートバンキング, 採用, 求人, 東京, 日本, リモート可, Payments UX, Money Movement, 決済・送金UX, ペイメントレール, multi-agent system design, agent orchestration architecture, human-in-the-loop approval gates" }