01
TRUST THE SIGNAL
A test suite is only valuable when engineers believe what a failure means.
SOFTWARE DEVELOPMENT ENGINEER IN TEST
3+ years building and scaling automation across UI, API, database and service-integration layers — with CI/CD, Docker and AWS at the execution layer.
Not vanity metrics. These are outcomes from problems I investigated, systems I changed, and automation I owned.
70% → ~5%
FALSE-FAILURE RATE
7+ HRS → <2 HRS
REGRESSION RUNTIME
0+
REGRESSION SCENARIOS
0
MICROSERVICES OWNED
ENGINEERING CASE STUDY / 01
How I helped turn an unreliable regression signal into one engineers could trust.
PROBLEM
Regression was producing more than 70% false failures during release windows.
A large regression suite was being run by multiple squads during a 2–3 day release sign-off window. The result was a huge amount of time spent investigating failures that were not actual product defects.
ROOT CAUSE
Continuous polling became a scalability problem.
Async events were being verified by querying MongoDB. The polling approach was reasonable for an individual test, but under concurrent execution across multiple squads it generated enough query volume to bottleneck the database.
INVESTIGATION
ENGINEERING CHANGES
01
Interval-based polling
Replaced continuous querying with controlled intervals, cutting query volume without losing async coverage.
Engineering decision I drove
02
Graceful failure behaviour
Made tests degrade predictably under load so a failure pointed at product behaviour, not infrastructure noise.
Engineering decision I drove
03
MongoDB → DocumentDB
Migration for additional headroom under concurrent execution.
Collaborative work with DevOps
FALSE FAILURES
>70% → ~5%
REGRESSION
7+ HOURS → ~2 HOURS
“The goal wasn’t simply to make the pipeline faster. It was to make a failed test mean something again.”
01
A test suite is only valuable when engineers believe what a failure means.
02
Good quality systems reduce dependence on people simply trying harder.
03
I prefer being involved from requirements and acceptance criteria through dev testing, rather than receiving finished stories and simply writing tests.
ENGINEERING / AUTOMATION SYSTEM
From test definition to pipeline execution, I work across the layers that make automation reliable at scale.
Scenarios written as readable behaviour, versioned alongside the code they cover.
UI and API layers share structure so tests stay maintainable as the product changes.
Generated data and typed deserialization keep parallel runs independent and assertions precise.
Containerised runs scale out per feature file so regression finishes inside a release window.
Artifacts, logs and device coverage handled by managed infrastructure rather than local machines.
Post-run capture makes a failure explainable and reruns targeted instead of wholesale.
RELIABILITY / DEVELOPER PRODUCTIVITY
Built a post-run failure capture and triage workflow that identifies failures and makes rerun and debugging more targeted.
Read case study →SCALABILITY / CI
Designed a pod-per-feature-file execution model with autoscaling that helped reduce Booking microservice regression runtime from 7+ hours to under 2.
Read case study →TEST ARCHITECTURE
Introduced dynamic test data generation and typed POJO deserialization using Java Faker and ObjectMapper.
Read case study →QUALITY ENGINEERING
Participate from requirements and acceptance criteria through development testing rather than treating QA as a downstream phase.
Read case study →FEATURED EXPERIMENT
An automation project using Playwright + TypeScript with Page Object Model, session reuse and scheduled GitHub Actions execution.
CURRENTLY EXPLORING