Shad Abbaslou / Contexts
● Open to global roles
~/overview
9 FILES · 5 RUNNING PRODUCTS
FIG. 01 · THE TOKEN FIELD LDN --:--

Shad Abbaslou · Product Design Lead · Hands-on across AI, Agents and Workflows

Design that
proves itself.

I lead product design on an enterprise AI programme: five product teams, and the design system nine designers build on. Ten years shipping enterprise software across AI, data platforms, healthcare and travel.

3,396 points  one per token in the system. Case file 001 Five products  rebuilt, running, and open from this page

[ The case files ]

What are you hiring for?

Five projects, each with its product rebuilt and running inside the file. Choose what the role needs and the files that show it are marked.

  • AI and agents 001, 002
  • Design systems 001, 002
  • User research 002, 003
  • Data and analytics 002, 004, 005
  • Healthcare and mobile apps 003
  • Enterprise tools 001, 002, 004, 005

001Storybook

Open InnovationProduct Design Lead · 2026

A design system that remembers why

~15s from a sentence to a working prototype
18 min read · 3,396 tokens · 91 rules ·

001Open Innovation

The design system of an enterprise ML platform, which five product teams and nine designers build on. I built it end to end with the frontend team, with a pipeline that turns one typed sentence into a working prototype.

Inside the file
  • The portal and the real Storybook, running
  • Tokens and numbered rules, and why dark mode cost no component code
  • A memory of the decisions that people and AI agents both read
  • The AI prototyping pipeline, step by step

Hiring forAI and agentsDesign systemsEnterprise tools

Open the case study: A design system that remembers why
002

Lead Product Designer · over three years

Three years inside a data platform

4 of 5 participants rated the redesigned modelling task easy
20 min read · two pods · one designer

002Astrato

The only designer, across both delivery pods, on a Snowflake-native analytics platform where data teams turn their own warehouse into models and dashboards. I redesigned the join canvas where every model begins, and designed Insights, an AI assistant that shows its SQL.

Inside the file
  • The workbook builder, rebuilt and running, to try
  • The join canvas, from a backlog ticket to a tested redesign
  • Moderated usability testing with attention heatmaps
  • Insights, the AI assistant, and the handoff to engineering

Hiring forData and analyticsAI and agentsUser researchDesign systemsEnterprise tools

Open the case study: Three years inside a data platform
003Phone apps

Product Designer · a year, discovery to beta

NHS mental health care, moved home

94% of a live NHS beta said it should continue
24 min read · 3 apps + clinician tool · 1 real clinic

003NHS

When COVID moved an NHS trust’s mental health care out of clinics and into homes, I took the service from discovery to a live beta: a private diary for teenagers, a summary for parents and one clinician tool for calls, with a simple app for older adults proposed on the same model.

Inside the file
  • The phone apps and the clinician dashboard, running
  • The research: interviews, a survey and co-design workshops
  • A privacy model built on consent, where sharing is always a choice
  • Testing, and a three-week beta in a real clinic

Hiring forHealthcareUser researchMobile apps

Open the case study: NHS mental health care, moved home
004

Novo NordiskLead Product Designer · built on Astrato

The number that was true twice

Reconstruction · no outcome claimed
18 min read · finance and supply planning

004Novo Nordisk

A reporting layer for Novo Nordisk’s finance and supply planning, built on Astrato. One quarter’s sales were up and down at once, depending on a basis no screen showed, so every figure now carries the basis it was worked out on.

Inside the file
  • The reporting cockpit, rebuilt with fictional data, to try
  • Why sales grew 32% and fell 4% in the same quarter, and both were true
  • Modelling across five systems that cannot change

Hiring forData and analyticsEnterprise tools

Open the case study: The number that was true twice
005

JBSLead Product Designer · built on Astrato

The spread, and what it is made of

Reconstruction · no outcome claimed
16 min read · plant and commercial team

005JBS

A spread board for JBS’s plant controller and commercial team, built on Astrato. The margin was decided in three systems on three clocks, and nobody saw it whole until a lot closed. The board puts it in one place.

Inside the file
  • The spread board, rebuilt with fictional data, to try
  • Three decisions in three systems on three clocks, drawn as one line
  • The margin, followed from the pen to the port and back into the price

Hiring forData and analyticsEnterprise tools

Open the case study: The spread, and what it is made of

How every file reads

Each one opens on its product, running, so you can use the thing before you read about it. Then the same path every time: the brief and what I owned, what moved, the outcome against the bar it was set, and what I would do differently.

Every file can be skimmed in about half its reading time.

[ Internal products ]

Three things I built for the team at Open Innovation and used inside the company: a deck, a class and a story site. Each one runs in the browser.

a·1Internal product

Open InnovationInteractive deck

The UX Files

Twenty production UX problems, each rebuilt broken so you can feel it, then flipped to the fix

44 slides · runs in the browser

a·1The UX Files

Twenty real UX problems from a production ML platform, each rebuilt broken so you can feel it, then flipped to the shipped fix.

Inside the deck
  • Every problem playable: the broken version first, then the fix that shipped
  • Navigation, tables and filters, spacing, states and information design
  • Each fix with its why, how and what
  • Why each fix stays fixed: numbered rules and an AI verifier

TaggedInternal product

Open the deck: The UX Files
a·2Internal product

Open InnovationInteractive class

Everyone Builds

The class that taught atomic design to the whole company, one brick at a time

6 labs · runs in the browser

a·2Everyone Builds

The interactive class that taught atomic design to the whole company: six playable labs, and one robot that refuses Comic Sans.

Inside the class
  • Atomic design taught with bricks: tokens, atoms and molecules snapped together live
  • The two ways the company prototypes with the design system
  • The AI playground that was built, then retired for something better

TaggedInternal product

Open the class: Everyone Builds
a·3Internal product

Open InnovationStory site

From Chaos to Canon

The design system’s story, built with the system’s own tokens

9 sections · runs in the browser

a·3From Chaos to Canon

The design-system story, built with the system’s own tokens, including a before and after you can drag, two years of UX debt apart.

Inside the site
  • Atomic design taken literally: each layer built only from the one below
  • Every visual decision through a three-layer chain of tokens
  • The memory that records why each rule exists, for people and AI agents
  • The waves of change that built it, each scoped, checked and logged

TaggedInternal product

Open the site: From Chaos to Canon

Five case files, and all five products rebuilt and running inside them: the NHS service with its three apps and its clinician tool, the Astrato workbook, the design system portal with its real Storybook, and two reporting boards built on Astrato for its clients. Open a file and use the software. Nothing here is a mockup.

Scope
Five product teams on one programme
Platform
The design system nine designers build on
Depth
The only designer for 29 people across two delivery pods, over three years
Hands-on
Design to code: every product here is built by me
  • ~15sfrom a sentence to a working prototypecase file 001
  • 4 of 5participants rated the redesigned modelling task easy, on tapecase file 002
  • 94%of a live NHS beta said it should continuecase file 003
  • +18%conversion on group bookings of ten or morea launch report

[ Core threads of my work ]

  1. 01Systems that hold3,396 tokens, 91 machine-checked rules, 231 components, 2,082 Storybook stories in the real Storybook
  2. 02Three years in one productAstrato: two delivery pods, one designer
  3. 03Research with real participants19 interview sessions, 70 survey responses, 3 workshops and a 21-person NHS beta, and three apps with a clinician tool built from it
  4. 04Hands on the codebaseEvery product on this site is one I built to prove the case
The five products, rebuilt and running here, stood up by your scroll. Hold one and it lifts.
booting…
● Live product · open the file to use it

A design system that remembers why

A token-governed system with numbered rules, an institutional memory read by humans and AI agents alike, a plain-English prototyping pipeline, and the real Storybook, 2,082 Storybook stories of the shipped components, running here, all for an enterprise ML platform.

Role
Product Design Lead
Company
Open Innovation
Timeframe
2026 · 100+ shipped releases
Built with
The frontend team
  • 3,396 CSS custom properties resolve through a three-layer chain, which is why dark mode cost zero component code.
  • 91 numbered rules and an append-only decisions log: 378 documents, 6,683 cross-links, read by people and machines.
  • Prototyping loop of ~15 seconds per iteration, designers and PMs build with real components by typing plain English.
Open & use the portal The Storybook Live story ↗
booting…
● Live product · open the file to use it

Three years inside a data platform

Both delivery pods of a Snowflake-native analytics platform, over three years of continuous delivery, from the join canvas to an AI assistant that shows its SQL.

Role
Lead Product Designer
Client
Astrato, by Vizlib
Duration
Over three years
Scope
Both pods · three years
  • The join canvas redesign started from a real backlog ticket (VK-4532): joins now route around tables, connect from any side, and take 4 clicks to create.
  • Moderated UserTesting sessions with attention heatmaps; four of five participants rated the redesigned task easy.
  • Two pods, one designer: data preparation (team of 9) and data visualization (team of 20).
Open & use the workbook Read the deck
booting…
● Live product · open the file to use it

Mental health care, moved home

COVID sent the clinicians home and the people they cared for were already there. A private diary for teenagers, a summary for their parents, a simple app for older adults, and one clinician tool made for calls rather than rooms. Research-led, and built to the GDS Service Standard.

Role
Product Designer
Client
NHS trust (anonymised)
Duration
A year, discovery → beta
Framework
GDS Service Standard
  • 19 interview sessions with 12 people across the three corners; surveys of 30 clinicians and 40 young people & parents.
  • Three co-design workshops, 12 clinicians in one, 8 teens in another, plus a privacy design that handled Gillick competence and parental consent.
  • Beta pilot in one clinic, 21 people over three weeks. The beta report: 94% of users said it should continue in use, a share given without its base.
Use the phone apps Use the dashboard Read the deck
booting…
● Live product · open the file to use it

The number that was true twice

A reporting layer for a company whose sales both rose and fell in the same quarter, where the honest answer depends on a basis nobody was showing.

Role
Lead Product Designer
Client
Novo Nordisk
Platform
Astrato on Snowflake
Status
Reconstructed
  • No measure renders without the basis it was computed on, and changing basis restates the whole screen at once.
  • Five validated source systems that cannot be altered, so the logic sits in a semantic layer above them rather than inside any of them.
  • A segment restructure that does not restate cleanly is drawn as a seam, not smoothed into one continuous line.
Open & use the cockpit Read the case file
booting…
● Live product · open the file to use it

The spread, and what it is made of

A reporting layer for a company that turns over BRL 417bn and keeps about 2.3% of it, where the margin is decided in three places that could not see each other.

Role
Lead Product Designer
Client
JBS
Platform
Astrato on Snowflake
Status
Reconstructed
  • Every lot carries a settled proportion: how much of its value is actually sold against how much is inventory held at an assumption.
  • The model keys on cut and destination, because that is where the price is decided, and rolls up to plant from there.
  • An export approval sits on the margin screen, because losing one redirects cuts to a cheaper market the same day.
Open & use the spread board Read the case file

The UX Files

Twenty real UX problems from a production ML platform, each rebuilt broken so you can feel it, then flipped to the shipped fix.

Format
Interactive deck · 44 slides
Problems
20, root-caused
Open live ↗

Everyone Builds

The interactive class that taught atomic design to the whole company: six playable labs, and one robot that refuses Comic Sans.

Format
Interactive class · 6 labs
Audience
The whole company
Open live ↗

From Chaos to Canon

The design-system story, built with the system's own tokens, including a draggable before/after two years of UX debt apart.

Format
Story site · 9 sections
Built
With its own tokens
Open live ↗

The delivered artifacts

Sixteen pages of the real thing: the sites, decks and classes as they were handed over. Not write-ups about the work; the work.

Five spines, one per artifact in the list below. Pull one out.
Where to go next
Open case file 001 → Every file and product
Contact
LinkedIn ↗ Read the CV

Last updated