Skip to content

(00) Introduction

Open to full-time, part-time and freelance

I design products and ship them.

See recent builds

I run the , draw the , design the , write the front to back and build the AI into it, then take it to production. One pair of hands and a team of agents, so nothing gets lost between deciding and shipping.

Currently
Product & Design Engineer, cMatter
Previously
baraka, Copper, Cognizant, Publicis Sapient
Open to
Full-time, part-time, freelance

Disciplines: Product design, Full-stack engineering, AI products, Coding agents, Design systems, Shipping to production.

(02) Expertise

Three disciplines, one overlap.

Most teams split these across three people and lose something at every handover. I hold all three.

01Research & interaction

Research

Talking to the people who will actually use the thing, mapping the flows, and settling how every state behaves before a line of production code exists. Keyboard paths and contrast get decided here, while they are still cheap to change.

Inputs
Users, constraints
Output
Flows, prototypes
Standard
WCAG 2.2
  • UX research
  • Interaction design
  • Prototyping
  • Accessibility

02Interface & systems

Interface

Tokens, components, typography and motion, plus the loading and empty states nobody remembers to draw. Built as typed, versioned code so the system is something engineers use rather than something they translate.

Tokens
Themed, versioned
Components
Typed APIs
Docs
Storybook
  • UI design
  • Design systems
  • Motion
  • Typography

03Production engineering

Production

Building the whole thing and taking it live: React and Next.js on the front, typed APIs and data behind it, and the AI features (agents, retrieval, tool use) wired in properly. Coding agents do the legwork; I set the architecture, review every change and hold it to the performance, testing and security bar real traffic needs.

Stack
Next.js, Node, Postgres
AI
LLMs, RAG, agents
Proof
TDD, Playwright, CI
  • Full-stack TypeScript
  • APIs & data
  • AI features
  • Coding agents
  • Performance

04The overlap

Research + Interface + Production =

the job.

Holding all three means the research survives into the interface, and the interface survives into production. Nothing is lost in translation, because nothing is translated.

Years shipping
0+
Projects delivered
0+
Sectors
0
Markets
0+

Sectors served: Fintech · Crypto · Energy tech · Enterprise AI · E-commerce · Automotive · Retail · IT services

(03) Component library

Small details, built properly.

Six components built for this site and documented the way I hand them to a team: a live preview, the decisions as values, the code, and the accessibility contract.

Components / Actions

Hold to confirm

Stable

A destructive action that cannot fire by accident. Press and hold to delete.

Press and hold

(04) How I work

From the first interview to the release.

Five stages, one pair of hands, so nothing gets lost between deciding and building.
  1. 01 / 05

    Listen

    Interviews and data until the real problem is obvious.

    What comes out of it

    • Interview notes and synthesis
    • A problem statement everyone signs
    • Success measures agreed up front

    Taste is the bottleneck.

    Agents can draft ten versions of a screen before lunch. None of them knows which one is right; that comes from sitting with the people using it.

    Example artefact

    Interview notes, round one

    Checks the total twice before sending
    Wants to see fees before, not after
    Screenshots receipts “just in case”
    Gave up when verification failed silently

    Synthesis

    “I don’t trust the total until I see where it came from.”

    Trust86%
    Clarity74%
    Speed52%

    Share of interviews that raised it

  2. 02 / 05

    Map

    Flows, states and edge cases, with the failure path first.

    What comes out of it

    • User flows with every branch
    • State inventory: empty, loading, partial, error
    • Keyboard and focus order

    Finish what you draw.

    A mockup is a proposal. The real work is what it left out: the loading state, the error nobody planned for, the connection that drops halfway through.

    Example artefact

    Onboarding flow: failure paths drawn first

  3. 03 / 05

    Shape

    Interface, type and motion, tested against real content.

    What comes out of it

    • High-fidelity screens with real numbers
    • Motion specs: curves, durations
    • Prototypes for the risky moments

    Feel is a specification.

    Spacing, rhythm, easing and focus order are not decoration. They separate software people tolerate from software they trust, so they get written down.

    Example artefact

    Available balance

    $12,480.50

    SendRequest
    2028 / 32
    −3% tracking
    44px target
    8px gap
  4. 04 / 05

    Systematise

    Tokens and components as typed, versioned code.

    What comes out of it

    • Design tokens in code
    • Components with documented states
    • Accessibility contract per component

    Accessible from the sketch.

    Contrast, keyboard paths and semantics are settled while it is still a drawing. Retrofitting them later costs several times more and never comes out as clean.

    Example artefact

    Colour tokens

    • color.ink#161616
    • color.paper#FAFAF9
    • color.line#EAEAE7
    • color.accent#C6F432
    • color.danger#C93A2E
    • color.ok#1E7A50

    Type and space

    Display28px
    Title20px
    Body15px
    Caption12px
    481216243248
  5. 05 / 05

    Ship

    The whole app in production, then measured and tuned.

    What comes out of it

    • Front end, APIs, data and AI features
    • Tests and CI on the journeys that matter
    • Performance, security and accessibility budgets

    A design that stops at the mockup is a proposal.

    Somebody still has to decide what happens on a slow connection. I would rather it was the person who designed it.

    Example artefact

    Release checklist

    • Every state designed and built
    • Keyboard path end to end
    • axe: no violations
    • Reduced motion respected
    • Playwright journeys green
    • Copy reviewed in context

    Performance budget

    • LCP< 2.5s
    • INP< 200ms
    • CLS< 0.1

    Budgets set before the build, checked in CI and after launch.

(05) Working with agents

Ten attempts. One decision.

Claude Code, Codex and Cursor compress the middle of the work, drafting ten versions of a screen in the time it took to draw one. Sitting with a user and deciding which version earns its place stays with me.

The pipeline, live

One change, from interview to release. Agents compress the middle; the first and last steps stay human.

  1. 01

    Discovery

    Sit with users, frame the real problem.

    Human

  2. 02

    Flows

    Map the paths, the states, the edge cases.

    Human

  3. 03

    Options

    Agents draft directions in parallel.

    Agents, supervised

  4. 04

    Judgement

    I choose the direction and the detail.

    Human

  5. 05

    Build

    Agents implement, I set the architecture.

    Agents, supervised

  6. 06

    Test

    Agents run the suite and review diffs.

    Agents, supervised

  7. 07

    Ship

    I check every state and take it live.

    Human

(06) Stack

A table of tools, sharpened by use.

66 tools. The ones in lime are what I reach for first; the rest are the supporting cast.
  • 1UdUI Design
  • 2IdInteraction Design
  • 3DsDesign Systems
  • 4UrUX Research
  • 5WiWireframing
  • 6PrPrototyping
  • 7DtDesign Tokens
  • 8TyTypography
  • 9MdMotion Design
  • 10RdResponsive Design
  • 11W2WCAG 2.2
  • 12ReReact
  • 13TsTypeScript
  • 14NjNext.js
  • 15RmRemix
  • 16AsAstro
  • 17TcTailwind CSS
  • 18CmCSS Modules
  • 19ScStyled Components
  • 20MoMotion
  • 21StStorybook
  • 22NdNode.js
  • 23PsPostgreSQL
  • 24NsNestJS
  • 25TrtRPC
  • 26GqGraphQL
  • 27PiPrisma
  • 28DrDrizzle
  • 29RiRedis
  • 30TqTanStack Query
  • 31ZuZustand
  • 32CcClaude Code
  • 33CuCursor
  • 34VaVercel AI SDK
  • 35CoCodex
  • 36LgLangGraph
  • 37OaOpenAI
  • 38AnAnthropic
  • 39OwOpen-weight LLMs
  • 40LlLocal LLMs
  • 41RaRAG
  • 42VdVector DBs
  • 43McMCP
  • 44EvEvals
  • 45ViVitest
  • 46PlPlaywright
  • 47TlTesting Library
  • 48CyCypress
  • 49VrVisual regression
  • 50VtVite
  • 51TuTurbopack
  • 52NxNx
  • 53TbTurborepo
  • 54FiFigma
  • 55FrFramer
  • 56DoDocker
  • 57AwAWS
  • 58VeVercel
  • 59GaGitHub Actions
  • 60AcAccessibility
  • 61CwCore Web Vitals
  • 62TdTDD
  • 63CdCI/CD
  • 64MnMonorepos
  • 65OsOWASP
  • 66DqDesign QA

(07) Experience

13 years of commits.

$ git log --career --oneline

8 commits · 3 tags · since 2013

  1. tag: v3Owning the product

  2. tag: v2Regulated, and at scale

  3. tag: v1Learning both sides

  4. initial commit

(08) About

Most products have a seam where design stops and engineering starts. I’ve spent 13 years working both sides of it so it doesn’t show.

  1. 01
    The seamWhy it matters

    You can feel the seam: the state nobody drew, the animation that snaps, the error message written by whoever reached it last. Closing it is less about talent than about never handing the work over halfway.

  2. 02
    The pathFreelance, agencies, startups, then regulated scale

    I started as a freelance front-end developer, moved into design through agency and startup work, and have since shipped for banks, a broadcaster, a crypto custodian, a trading app and an AI platform. Five of those years were in London.

  3. 03
    The work nowcMatter

    Today I design, build and ship the product at cMatter, an agentic AI platform for buildings and energy: the research, the interface, the full-stack code and the AI underneath, with coding agents doing the legwork and me making the calls. I work in English and Arabic.

Building products
since 2013

Medhat Albsugy

Product & Design Engineer, AI-native

Experience
13+ years, 5 in the UK
Domains
Fintech, digital assets, AI products
Builds with
Design, full-stack TypeScript, AI
Languages
English, Arabic
Open to full-time, part-time and freelance

(09) Contact

Let’s build something that ships.

Open to full-time roles, part-time engagements and freelance projects. If you need a product designed, built and shipped, with AI inside it or behind it, that is the work I do.

Click ↵ to email me.

Medhat Albsugy

© 2026

Designed and built with coding agents, three.js, Motion and Lenis.