Skip to evidence →
tanyasingh
← Back to evidence index
Exhibit B · Open

Evidence folder

Jira Accessibility Compliance

Client
Atlassian Jira · via EPAM
Role
Accessibility SME / Lead Engineer
Timeline
EPAM · remediation program
70%Violations resolvedAcross the prioritized remediation backlog for Jira surfaces.
82%Axe coverageAutomated checks catching most regressions before production.

Objective

Drive WCAG 2.1 AA remediation across a massive React surface area used by millions of daily users — and leave behind governance so regressions die in CI, not production.

Decision log

Choices under review — why they stuck, and what they cost.

  1. Decision 01

    Prioritize high-traffic violation classes over one-off polish

    WhyWith 2,000+ components in play, triage had to follow user impact and repeatable fix patterns.

    TradeoffSome edge components waited — core journeys and reusable primitives moved first.

  2. Decision 02

    LLM-assisted ARIA drafts with human review

    WhyRepetitive remediation burned senior hours; Copilot-assisted drafts with SME review compressed cycle time without abandoning judgment.

    TradeoffRequired clear review rituals so speed never outran correctness.

  3. Decision 03

    Axe gates in CI as non-negotiable

    WhyManual audits alone cannot hold the bar on a living product.

    TradeoffPipeline noise early on — tuned rules until signal beat friction.

What we built

  • Remediation program across 2,000+ React components
  • LLM-assisted ARIA fix playbooks with human review
  • Axe-integrated CI quality gates

Filed stack · React · ARIA · Axe · CI/CD · GitHub Copilot · WCAG 2.1 AA

Panel observations · people

  • Obs. A

    Codified remediation patterns other engineers could execute without waiting on a single SME

  • Obs. B

    Paired with teams to teach semantics and keyboard models, not just ticket dumps

  • Obs. C

    Turned accessibility from a late audit into a shared quality gate

Discuss this kind of engagement → Contact