/

Calm principles for high-stakes software.

Six principles that guide how I approach complex enterprise products—from information hierarchy and resilient states to accessibility, design systems, and responsible AI use.

Explore 4 Pillars
Principles
06
Applied in product work
Accessibility
WCAG 2.1 AA
Keyboard, structure & states
Design Tokens
W3C DTCG
Semantic token workflows
AI Practice
Human-Reviewed
Human accountability

Reducing complexity before designing screens

Before designing screens, I first understand the workflow, information relationships, user priorities, and decision points. The interface comes after the system makes sense.

(01) PrincipleSystem Modeling

Understand the system before drawing screens

Enterprise problems are systemic. Designing isolated screens without modeling workflows, states, and data contracts produces brittle software.

Before sketching or opening Figma, I map entity relationships, operator mental models, permissions, and asynchronous boundaries. When the underlying model is clear, the interface practically designs itself.

  • Map data flows and edge states before UI layouts
  • Model operator goals rather than surface decoration
  • Treat system constraints as generative design inputs
(02) PrincipleCognitive Hierarchy

Reduce complexity through hierarchy

High-density interfaces must not feel chaotic. Visual hierarchy, clear contrast, and deliberate typography restore cognitive calm.

Operators handle high-stakes decisions under time pressure. The layout must guide the eye instantly: primary status first, secondary context on demand, and destructive actions protected by clear confirmation.

  • Prioritize cognitive ergonomics over decorative whitespace
  • Use typographic scale to establish instantaneous scanning paths
  • Protect critical user actions with unambiguous signifiers
Example · Information Hierarchy

A three-level information hierarchy

Designed for faster scanning and prioritisation

One way I structure dense workflows is to separate immediate status, contextual detail, and guarded actions so users can understand priority without every element competing for attention.

Tier 1 · Immediate
Primary Status

The most important state, status indicators, and key values receive the clearest hierarchy for quick recognition.

Goal: immediate recognition
Tier 2 · Contextual
Details & History

Supporting information, history, and secondary metadata remain available without overpowering the primary task.

Goal: on-demand inspection
Tier 3 · Guarded
High-Impact Actions

Destructive or high-impact actions use clear consequences, confirmation, and deliberate friction before they are completed.

Goal: reduce accidental destructive actions

Resilient states and accessible interaction

Responsive behavior, empty and error states, keyboard interaction, focus behavior, multilingual layouts, and assistive-technology needs should be designed as part of the product—not added at the end.

(03) PrincipleBoundary States

Design for states, not screenshots

Real software lives in loading latencies, network interruptions, partial data sets, and error recoveries.

Static mocks hide operational reality. A robust design system accounts for empty states, asynchronous delays, boundary overflows, and resilient recovery paths as first-class specifications.

  • Every component specifies empty, loading, error, and overflow states
  • Eliminate optimistic illusions that disguise backend latency
  • Test layouts against extreme multilingual and localized data lengths
(04) PrincipleAccessibility Baseline

Build accessibility into the system

Accessibility is not an audit checklist at launch. It is an architectural foundation engineered directly into core tokens and components.

Meeting WCAG 2.1 AA requires structural semantic HTML, high-contrast ratios, visible focus indicators, screen reader landmarks, and keyboard navigation engineered from the first token.

  • WCAG 2.1 AA baseline verified during component creation
  • Keyboard-first navigation with focus trapping where appropriate
  • Tokens enforce contrast ratios automatically across dark and light palettes
Interactive Example · Bidirectional LayoutPrinciples 3 & 4

Explore how Arabic RTL changes layout direction while mixed-direction numbers and interface semantics remain readable.

Bilingual RTL & LTR Mechanics Engine

Bidirectional Cognitive Ergonomics & Layout Mirroring · Arabic & English Enterprise Workflows

Material Requisition Order

Central Operations & Logistics Depot

REF: 4810-A
Tariq Al-Mansoor
Regional Supply Chain (GCC-4)
SAR 3,850.00
Inventory data & compliance verified
1. Reading Direction & Scanning

Primary eye scan flips from right-to-left while retaining numeric data order, preserving the natural cognitive reading pattern across Arabic dashboards.

2. Numerical Formats & Data

Numerals, phone numbers, and currency markers adapt to localized regional conventions while tabular data alignment preserves numerical hierarchy.

3. Directional Semantics (Fitts's Law)

Forward navigation chevrons, progress steppers, and submit actions reverse seamlessly to indicate forward progression in RTL without user friction.

Design tokens that connect design and implementation

Shared token conventions, component names, states, and implementation rules reduce handoff friction and keep design intent clearer as products scale.

(05) PrincipleDesign Tokens & Code

Treat design and code as one workflow

When design and engineering share tokens, components, and mental models, handoff friction disappears.

I work across Figma variables and React/TypeScript codebases. Using W3C DTCG-aligned design tokens ensures that design changes translate into production code with clarity and confidence.

  • Shared token contracts between Figma and frontend codebases
  • Storybook documentation acts as the living source of interface truth
  • Fewer handoff meetings, faster iteration, zero translation ambiguity
Interactive Example · Token HierarchyPrinciple 5

See how primitive values, semantic intent, and component-level decisions can be separated so themes and product changes remain easier to manage.

W3C DTCG Token Hierarchy Compiler

Interactive token inheritance simulator demonstrating 3-tier token resolution.

Compilation Pipeline (Atlas Blue)3-tier resolution: global, contextual, scoped
01Primitive
Token Identifier
color.palette.atlas.500
Resolved Value
#1677A8RAW HEX

Raw foundation hex value decoupled from semantic intent.

02SemanticInherits from 01
Token Identifier
color.action.primary.default
Resolved Value
{color.palette.atlas.500}

Contextual intent alias referencing the primitive tier.

03ComponentInherits from 02
Token Identifier
button.primary.background
Resolved Value
{color.action.primary.default}

Scoped contract consumed directly by production React components.

Live Component Resolution
Production CSS

Real-time component consumption. The button binds to the Component Tier contract, decoupling React layout from raw primitive palette values.

CSS Custom Properties
:root {
--button-bg: #1677A8;
}
Resolved Fill:#1677A8
Target Element:
W3C DTCG Format Export
{
  "button": {
    "primary": {
      "background": {
        "$value": "{color.action.primary.default}",
        "$type": "color",
        "$description": "Primary action button fill"
      }
    }
  }
}

Human-reviewed AI in product work

I use AI to accelerate exploration, synthesis, repetitive analysis, and prototyping while keeping problem framing, accessibility review, ethical judgement, and shipping decisions with people.

(06) PrincipleResponsible AI

Use AI selectively

AI assists, people decide. Technology should accelerate exploration, not replace critical human judgment.

I leverage AI tools to synthesize research transcripts, generate realistic test data, explore token transformations, and run preliminary heuristic checks. But human designers must own problem framing, ethics, accessibility verification, and shipping decisions.

  • AI accelerates exploratory tasks and data synthesis
  • Human ownership of design decisions, trade-offs, and ethics
  • No autonomous compliance claims or hallucinated certainty
Interactive Example · Human ReviewPrinciple 6

A disciplined approach to integrating AI into product craft without compromising human responsibility.

Responsible AI Framework · Division of Labor Architecture

A calibrated operational matrix: AI amplifies speed and synthesis, while humans own judgment, taste, and ethics.

Human-in-the-Loop Protocol Active
Select Operational Workflow Vector
Usability & Triage

Automated Heuristic Violation Scanning & Severity Triage

Accelerating interface triage against Nielsen Norman 10 heuristics and WCAG 2.1 AA rules without relying on generative guesses.

~80% faster initial classification
Machine Vector · AI Acceleration
Speed & Synthesis
Operational Computation

Scans DOM elements against 40+ structured usability heuristics and WCAG contrast standards (as built in AI UX Audit Lite).

Accepted Machine Inputs

DOM tree, viewport screenshot & bounding box coordinates

Machine Guardrail: Cannot fabricate issues or declare compliance; restricted to verifiable DOM heuristics.
Human Authority · Cognitive Stewardship
Taste & Accountability
Non-Delegable Human Decisions
  • Arbitrates between theoretical violation and real-world task context
  • Decides whether to resolve friction via persistent labels, tooltips, or flow re-architecture
  • Ensures critical operational errors (e.g. clinical or financial) are never obscured
Signing Threshold:100% human-verified severity triage before engineering sprint backlog
Operational Artifact Specimen: Clinical Trial Patient Dosage Form
Heuristic 01 · Visibility of System Status
01. Raw Vector Input
DOM Node: <button type='submit' class='btn-save'>Confirm Administration</button>
02. AI Processing Output
FLAG [P2 Medium]: Async network mutation lacks visual progress feedback. No inline state transition detected.
03. Human Action & Verdict

Approved finding. Add disabled state, spinner token, and aria-busy='true'. Do not hide button to prevent layout shift.

Outcome: Verified Fix · WCAG 4.1.2 & Heuristic 01 CompliantAudited & Signed by Aftab Khan
Engineering Guardrails & Ethical Axioms
Accessibility Standard
Zero Hallucinated Compliance

AI will never certify WCAG or regulatory compliance autonomously. Only physical keyboard navigation, screen-reader testing, and human verification constitute proof.

Architectural Integrity
Predictable Token Pipelines

Design token transformation is strictly schema-verified (W3C DTCG / Zod), preventing token drift between design decisions and production code.

Data Privacy
Zero-Retention Privacy

Enterprise client codebases, clinical workflows, and sensitive data are processed strictly in ephemeral, zero-retention memory. No client IP trains external models.

Ethical Accountability
Unconditional Human Authorship

The designer and engineer who ship the commit own 100% of the production reality. AI is a precision amplifier, but the human signs the deployment contract.

See how these principles show up in product work.

Explore representative work and career experience across healthcare, fintech, logistics, workforce products, design systems, accessibility, and frontend implementation.

View Career Experience