From Zero to 1,000+ Scripts: Embedded QE Practice at Torch, Leadership’s Platform of Choice
Torch connects leadership growth to business performance through expert coaching, AI guidance, and organisational intelligence — trusted by FICO, Twitch, Reddit, and TripAdvisor. We built their entire QE practice from nothing, and ran it.
Where leadership development meets platform engineering rigour.
Torch is a leadership development and coaching platform that connects leadership growth directly to business performance through three pillars: expert human coaching, AI-powered guidance via the Spark AI agent, and organisational intelligence that surfaces insights from coaching data. Trusted by FICO, Twitch, Reddit, and TripAdvisor, Torch works at the intersection of human expertise and intelligent automation. The platform spans a complex multi-service architecture including meeting services, assessment services, integration services, a RAG-powered AI interview system, video and audio interview workflows, and deep accessibility requirements. Each service requires independent validation at the UI, API, job, and database levels — making Torch one of the most architecturally complex QE engagements in the practice. Torch's deep commitment to accessibility — supporting NVDA and VoiceOver screen readers across Windows and macOS — combined with its rapid feature velocity and major dependency changes made this an engagement that demanded both deep technical skill and sustained operational consistency.
Starting from nothing, building 1,000+ automated scripts, shipping AI-integrated coverage, and owning the release sign-off — this is what a full QE partnership looks like.
At a glance.
Built, scaled, sustained.
The Torch engagement evolved from building a QE practice from scratch, through scaling into a 1,000+ script automation suite integrated across CI/CD pipelines, to a full AI-augmented QE operation covering accessibility, microservices, and AI-integrated features — with the team owning full release sign-off throughout.
Built, scaled, sustained.
The Torch engagement evolved from building a QE practice from scratch, through scaling into a 1,000+ script automation suite integrated across CI/CD pipelines, to a full AI-augmented QE operation covering accessibility, microservices, and AI-integrated features — with the team owning full release sign-off throughout.
No QE Process or Infrastructure
Zero QA infrastructure to start. No test cases, no feature matrix, no process, no automation. The team was brought in to build everything — and own it as an embedded function.
Rapid Development & Continuous UI Revamps
Major interface overhauls could invalidate large portions of the automation suite simultaneously. Keeping automation current with a rapidly changing UI required a continuous refactoring discipline alongside ongoing feature coverage work.
Complex Multi-Service Architecture
Assessment, meeting, and integration services each required validation at the UI, API, job, and database levels. AWS Lambda jobs and cross-service flows introduced dependencies that demanded test isolation and layered verification strategies.
Major Dependency Upgrades
A full Cypress version upgrade broke the majority of automation scripts simultaneously. Stabilisation required nearly a full month of focused work on an isolated branch — rebuilding and retesting the entire suite before it could be reintegrated.
From nothing to a 1,000+ script QE operation.
The team built Torch's entire QE capability from scratch — beginning with process and infrastructure, scaling into comprehensive automation, and maturing into a full AI-augmented QE operation with ownership of release sign-off and new joiner mentoring.
Built QE from Scratch
Complete QE infrastructure: feature matrix, requirements traceability matrix, test cases in Zephyr Scale, release cycle structures, Confluence documentation hub. The team owned the QE pod, signed off every release, and mentored new joiners into the practice.
1,000+ Cypress Automation Scripts
Automation coverage built to ~80% of the platform's test surface. Scripts integrated into CircleCI and GitHub Actions for daily sanity runs and weekly release regression cycles. Weekend cron jobs executed full regression with Slack-based failure reporting — keeping the team informed without manual intervention.
AI-Integrated Automation with Cursor AI
Cursor AI used throughout for writing, reviewing, and refactoring scripts — accelerating coverage velocity significantly. The team also implemented full automation of video and audio interview flows for Torch's RAG-powered AI interview application, extending QE into the AI feature surface.
Shift-Left Testing & Microservices Validation
Scripts developed in parallel with feature development to catch issues before they reached regression. AWS Lambda jobs and end-to-end flows validated across all three services — including database and job-level validation — providing multi-layer confidence on every release.
Dependency Upgrade Stabilisation
When a full Cypress version upgrade broke the majority of automation scripts simultaneously, the team executed a disciplined month-long stabilisation effort on an isolated branch. The entire suite was rebuilt, re-validated, and reintegrated — with no permanent regression in coverage.
Deep Accessibility & Cross-Platform Testing
Comprehensive accessibility testing using NVDA, VoiceOver, axe, and WAVE across Windows and macOS. BrowserStack for cross-browser and cross-device validation. Lighthouse and K6 for performance benchmarking. Customer support collaboration for production issue diagnosis and root cause identification.
The numbers.
Cypress Automation Scripts
Platform Coverage Automated
Manual Testing Effort Reduced
Recognised Directly by Leadership
What changed.
Automated Scripts Built
Covering ~80% of Torch's platform surface — assessment, meeting, integration services, AI features, and accessibility flows.
Test Surface Automated
Built incrementally across the engagement as features shipped — maintained through UI revamps and a full Cypress dependency upgrade.
Manual Effort Reduced
Automation absorbed the testing volume that would otherwise have required multiple additional manual testers across each release cycle.
Microservices Validated
Assessment, meeting, and integration services — each validated at UI, API, job, and database levels for full-stack confidence.
Automation Built Alongside Features
Scripts developed in parallel with development — catching defects at the earliest possible moment in the release cycle.
Multiple Appreciations from Leadership
Recognised directly by Torch's C-suite and engineering leads across the engagement — a reflection of the practice's impact on product quality and release confidence.
The stack.
Torch represents the fullest expression of what a long-term, embedded QE partnership can deliver. Starting from zero — no process, no tests, no documentation — the team built everything: 1,000+ automated scripts, AI-integrated test coverage, microservices validation, shift-left workflows, accessibility testing, and full pod ownership with release sign-off authority. The depth of this engagement, and the recognition it earned from every level of the organisation, speaks to what happens when quality engineering is treated as a first-class function rather than an afterthought.