Enterprise QE at full depth
1,000+ automated tests and end-to-end pod ownership. How a multi-year embedded QE engagement built a complete quality practice from zero — Cypress automation at 80% surface coverage, AI-integrated test development, microservices and accessibility testing, shift-left workflows, and full pod-level release sign-off.
Automated test cases
Automation coverage
Manual effort reduced
C-suite appreciations
Quality engineering for a leadership development platform.
Torch is a leadership development and coaching platform designed to unlock the potential of individuals, teams, and organisations. It connects leadership growth directly to business performance through expert human coaching, AI-powered guidance (the Spark AI agent), and organisational intelligence — helping companies lead through transformation, AI adoption, restructuring, and sustained resilience. Trusted by leading brands including FICO, Twitch, Reddit, and TripAdvisor, Torch serves HR teams, leadership, and coaching professionals across enterprise clients globally.
The platform spans a complex multi-service architecture — meeting services, assessment services, integration services, a RAG-powered AI interview system, video/audio interview workflows, and deep accessibility requirements — making QE a critical enabler of platform reliability and growth. We built Torch's QE capability from the ground up: a complete practice from nothing, scaled to 1,000+ automated tests, full pod-level release sign-off, and cross-functional recognition from CEO to customer support.
Build everything. Own everything.
Torch was a deeply complex, long-running engagement that required building a complete QE practice from nothing — across a fast-moving platform with frequent releases, major UI overhauls, a multi-service architecture, AI-integrated features, and strict accessibility requirements. The team didn't just join an existing QE function — they had to create one, own it, and prove its value at every layer: from Lambda jobs and microservices, to UI regression, to accessibility, to release sign-off.
No QE process or documentation
Started with zero infrastructure u2014 no test cases, no feature matrix, no RTM, no regression structure, no documented process. Everything had to be established from scratch while simultaneously testing an actively developed platform with frequent weekly releases.
Rapid development & continuous UI revamps
Frequent releases, continuous feature development, and significant UI revamps created constant pressure on both manual and automated coverage. Major interface changes could invalidate large portions of the suite, requiring disciplined maintenance and stabilisation after each overhaul.
Complex multi-service architecture
Multiple interdependent services u2014 assessment, meeting, integration u2014 each requiring validation at the UI, API, job, and database levels. When the meeting service UI was revamped, the full feature including related Lambda jobs and database state had to be re-validated end-to-end.
Major dependency upgrades
Cypress and dependency upgrades periodically broke the majority of automation simultaneously. A single major upgrade cycle took nearly a month to stabilise u2014 executed on an isolated branch to avoid contaminating the main suite, with careful engineering and DevOps coordination.
Foundation, automation, AI, ownership.
Built QE from scratch
Established the entire practice u2014 feature matrix, RTM, test cases in Zephyr Scale, release cycle structures, Confluence documentation. The team owned the QE pod, coordinated with development and product, signed off releases, and mentored new joiners (developers and QEs).
1,000+ Cypress automation scripts
Automated over 1,000 cases covering ~80% of the platform's test surface. Integrated into CircleCI and GitHub Actions for daily sanity and weekly release regression. Cron jobs ran over weekends with results posted to Slack and Cypress Cloud.
AI-integrated automation with Cursor AI
Cursor AI was used throughout the lifecycle to write, review, and refactor Cypress scripts. Additionally, AI-integrated automation was built for the Torch AI interview application u2014 a RAG-powered tool u2014 including full automation of video and audio interview flows within Cypress.
Shift-left automation approach
Implemented a shift-left model: automation scripts developed in parallel with feature development, ensuring coverage was ready at point of delivery rather than as a trailing activity. Significantly compressed feedback loops and reduced manual regression burden every cycle.
AWS and microservices testing
Testing extended into the cloud infrastructure layer u2014 validating AWS Lambda jobs and end-to-end flows across the assessment, meeting, and integration services. When the meeting service UI was revamped, the full feature was re-validated across UI, backend jobs, and database state.
Deep accessibility testing
Comprehensive accessibility testing using NVDA and VoiceOver for screen readers, axe and WAVE for automated checks, and manual testing across Windows and macOS covering colour gradients, themes, and contrast. Cross-browser via BrowserStack ensured parity across all environments.
QE across every layer of the platform.
The engagement covered the entire surface of the platform — not just UI, but Lambda jobs, microservices, accessibility, performance, AI features, and cross-functional collaboration with customer support. Each layer was treated as a first-class testing domain.
What the work delivered.
1,000+ automated test cases delivered in Cypress
Covering ~80% of the test surface u2014 with weekend cron, daily sanity, and on-demand pipelines in GitHub Actions and CircleCI, all reporting to Slack and Cypress Cloud.
Shift-left automation model implemented
Scripts delivered in parallel with feature development u2014 compressing feedback loops and reducing manual regression for every weekly release.
AI-integrated automation built for the Torch interview app
Full Cypress automation of video and audio interview flows and RAG-powered coaching use case testing, with Cursor AI used throughout.
AWS and microservices layer validated end-to-end
Lambda jobs, assessment, meeting, and integration services tested u2014 including database and job validation after the meeting service UI revamp.
Deep accessibility coverage delivered
NVDA, VoiceOver, axe, WAVE across Windows and macOS u2014 screen readers, colour gradients, themes, cross-browser parity via BrowserStack.
Multiple appreciations from CEO, CTO, dev and product leads
Recognition for contributions to platform reliability, process maturity, and cross-functional collaboration throughout the engagement.