Senior Software Engineer · Frontend + platform · London

The best platform work disappears into the product.

I’m Atheer Musa. I build design systems, developer tooling and AI-assisted workflows that help large teams ship one coherent, accessible experience. At Lloyds Banking Group, that means leading Cancara Labs and building shared semantic contracts and tooling for teams shipping across eight brands and React, AEM, iOS and Android.

View signature work Start a conversation

Product craft, developer experience and the systems between them.

Four delivery checkpoints

A trace from need to evidence.

HUMAN GATE

Intent

Turn the real need into an explicit source of truth.

Start with approved design, documentation and the decision a person is trying to make—not a tool looking for somewhere to be used.

OutputSource of truth

Current scope

One shared system. 8 brands. 4 platforms.

Shared semantic contracts and tooling for teams shipping React and AEM on web, and native iOS and Android. My hands-on work centres on web foundations, automation, testing, documentation and analytics.

Key automated brand workflowI reduced a multi-month manual effort to approximately half a dayComparison is against the previous repeatable manual rebrand path.

Testing + visual regressionI delivered improvements estimated to save 500 developer hours annuallyEstimate covers avoided component, theme and rebrand validation effort.

Signature systems

What I can show without showing the work.

Internal screens and code stay private. The architecture, decisions and evidence do not have to. These two systems show how approved design becomes a governed AI workflow, and how one semantic contract travels across brands and platform teams.

01 · Lloyds Banking Group · Cancara Labs

Developer workflow

From design intent to governed delivery

I lead and own Cancara Labs within CancaraOS, turning recurring design and engineering needs into AI-assisted workflows, reusable patterns and MCP server offerings, with evaluation and human review built in.

Explore the governed AI delivery system

Sanitised offering · Design-to-code evaluation

A governed evaluation path—not an autonomous shortcut.

01User + trigger
A design-system engineer selects an approved Figma node and target component
02MCP boundary
Approved Figma, documentation and repository capabilities only
03Candidate
The reusable workflow produces an implementation and evaluation record
04Evaluate
Structural comparison, visual diff and recorded failure modes
05Human gate
A named engineer accepts, rejects or iterates before delivery

Rejection loopRejected output returns to the workflow with its failure mode recorded.

Status: Owned MVP and evaluation foundations in active delivery
Attributable result: Moved evaluation from informal exploration into a repeatable, reviewable proof path

Representative current offering. Internal schemas, prompts, evaluation data and product code are intentionally omitted.

What I established

Moved AI and platform exploration from side-of-desk activity into owned MVPs, delivery plans and measurable proof paths.
Made evaluation, visual comparison and human approval part of the delivery architecture.

02 · Lloyds Banking Group · Cancara design system

Multi-platform system

Cancara: one system across eight brands and four platforms

I help maintain and evolve the shared Cancara foundations within CancaraOS, supporting teams shipping eight brands across React, AEM, native iOS and Android through semantic contracts, automation, testing, documentation and analytics.

Explore the Cancara design system Visit CancaraOS (opens in a new tab)

Sanitised contract topology

One semantic contract, resolved four platform-native ways.

Shared intent: Approved semantic token + explicit change contract

React: Token package + component API

AEM: Mapped theme + content implementation contract

iOS: Platform-owned native semantic resources

Android: Platform-owned native semantic resources

Evidence loop: Accessibility checks, visual regression and adoption analytics

The shared contract is mine to help shape; native implementation stays with the specialist platform teams.

Measured change

I reduced a multi-month manual effort to approximately half a day
I delivered improvements estimated to save 500 developer hours annually

Earlier product impact

The work that built the foundations: accessible product systems, monorepo delivery and customer-critical journeys.

Experience

Over five years of turning ambiguity into working systems.

Lloyds Banking Group

Senior Software Engineer

June 2024 — Present

Lead and own Cancara Labs within CancaraOS while maintaining web foundations and shared delivery contracts for a design system serving eight brands and four platforms.

Selected proofSet the vision for reusable AI workflows, implementation patterns, MCP servers and governed delivery offerings.

JATO Dynamics

Frontend Developer

March 2023 — June 2024

Led an accessibility-focused React library and token-driven design system spanning Figma, documentation and four SaaS products.

Selected proofAligned reusable implementation standards across several product teams.

Zego

Software Engineer

January 2021 — March 2023

Built customer onboarding and purchase journeys across React micro-frontends and Python-backed services.

Selected proofLed a 50+ component library used across the business.

Working principles

Standards I want the work to survive.

These are not values added after delivery. They are architectural constraints I use to decide what to build, what to automate and what evidence to leave behind.

01
Accessibility belongs in the component API.
If every product team has to remember to add it later, the platform has not done its job.
Seen inReact Aria and WCAG AAA foundations across four JATO SaaS products.
02
Human review is part of AI architecture.
Evaluation, visible failure states and a named decision point belong inside the workflow.
Seen inGoverned AI and MCP offerings within CancaraOS.
03
Platform metrics should change decisions.
Adoption data matters when it changes what the system team improves, documents or retires next.
Seen inAutomated Cancara analytics and product instrumentation at JATO.

About

I care about the parts people never see—and how they change what people feel.

I studied Electronic Engineering at King’s College London; my dissertation became an AR iOS application using facial recognition. That same curiosity now runs through design systems, developer tools, native prototypes and the homelab services I operate outside work.

I mentor colleagues through Coach2Code because good platforms are not only technical. They should help more people understand the system and do confident work inside it.

What comes next

I’m looking for teams where product craft and platform depth belong together.

If you’re building a customer experience, developer tool or applied AI system that has to earn people’s trust, I’d like to hear what makes it difficult.

[email protected]View GitHub (opens in a new tab)